as Quantum

Quantum Mechanics

Information
(and only

a little

more)

Christopher A.Fuchs
Computing Science Research Center

Bell Labs, Lucent Technologies

Room 2C-420, 600700 Mountain Ave.
Murray Hill, New Jersey 07974, USA

Abstract
In this

paper,

I
try

once

again to

cause some

good-natured trouble. The issue remains, when will we

ever stop

the taxpayer

burdening

with conferences

devoted to the quantum foundations?

expressed that

found

no end

to reduce

statements

quantum

of crisp

theory

physical

axiomatic) significance.

The suspicion is

will be in sight until

a means is
or three

to two

(rather

than abstract,

In this regard,

no tool appears

better calibrated

for

direct assault than quantum

a strained application of
a time-honored problem, this method
precisely because a large partbut not

theory. Far from

information

the latest fad to

holds promise

allof the structure of quantum theory

concerned

information.

has always

It is just that the physics

community needs reminding.

paper, though
core, corrects one

This
its

observations
particular,
mechanics
in the

beyond

mistake
the

one

Iidentify

and

traditionally

is

ascribed to

the

offers

previous

version.

element

that I
would not label

theoryit

as
sev-eral

taking quant-ph/0106166

of

In

quantum

a subjective term

integer

parameter

a quantum system

via its

Hilbert-space dimension.

Introduction
Quantum

as a
with us for

theory

structure has been

weather-sturdy

75

years now.

Yet, there is a sense in which the struggle for its

construction remains. I
say this because

one can
a year has gone by in the last 30
when there was not a meeting or conference
devoted
to some
aspect of the quantum
check that not

foundations.

Our meeting in V axj

o, Quantum

Theory: Reconsideration of Foundations,

is only

one in a long, dysfunctional line.

But how did this come about? What is
cause of this year-after-year sacrifice to
great

mystery?

for want of

the

it is, it cannot be

Whatever

a self-ordained

the

solution: Go to

any

meeting, and it is like being in a holy city in great

,
,

tumult. You will find all the religions

with all

their priests pitted inholy warthe Bohmians [3]

the

Consistent

Historians

[4],

the

Transactionalists[5], the Spontaneous Collapseans

[6]

the Einselectionists

Objectivists
and

[7]

[8], the outright

many more

the Contextual

Everettics

[9, 10],

beyond that. They all declare to

see the light, the ultimate light. Each tells us that

if we will accept their solution as our savior,
then we too will see the light.
1

This

viewed

paper,

as a

though

Details of the changes

present

paper,

substantially

continuation

and

can be found

Section 1
1

longer,

amendment

should

to

Ref.

in the Appendix

Substantial

further

be

[1].

to the

arguments

defending

transition

stance implicit
stance implicit

from

the

objective

in Ref. [1] to the subjective

here

can be found

Bayesian
Bayesian

inRef. [2].

1
A Fraction of the Quantum Foundations Meetings since 1972

1972

The Development of the Physicists Conception of Nature, Trieste, Italy

1973

Foundations of Quantum Mechanics and Ordered Linear Spaces,

1974

Quantum Mechanics,

1975

Foundational Problems inthe Special Sciences, London, Canada

1976

International Symposium

1977

International School of Physics Enrico Fermi, Course LXXII:

1978

Stanford Seminar

1979

Interpretations and Foundations of Quantum Theory, Marburg, Germany

1980

Quantum Theory and the Structures of Time and Space, Tutzing, Germany

1981

NATO Advanced Study Institute

Marburg, Germany

a Half

Century Later, Strasbourg, France

on Fifty

Years of the Schr odinger Equation,

Vienna, Austria

Problems inthe Foundations of Physics, Varenna, Italy

on the Foundations

of Quantum Mechanics, Stanford, USA

on Quantum

Optics, Experimental

Gravitation, and Measurement Theory, Bad Windsheim, Germany

a Tribute to Louis

1982

The Wave-Particle Dualism:

1983

Foundations of Quantum Mechanics inthe Light of New Technology,

de Broglie, Perugia, Italy

Tokyo, Japan

1984

Fundamental Questions inQuantum Mechanics, Albany, New York

1985

Symposium

on the Foundations

of Modern Physics: 50 Years of

the Einstein-Podolsky-Rosen Gedankenexperiment, Joensuu, Finland

1986

New Techniques and Ideas in Quantum Measurement Theory, New York, USA

1987

Symposium

on the Foundations

of Modern Physics 1987: The Copenhagen

Interpretation 60 Years after the Como Lecture, Joensuu, Finland

1988

Bells Theorem, Quantum Theory, and Conceptions of the Universe,

1989

Sixty-two Years of Uncertainty: Historical, Philosophical and

1990

Symposium

Washington, DC,USA

Physical Inquiries into the Foundations of Quantum Mechanics, Erice, Italy

on the Foundations

of Modern Physics 1990: Quantum Theory of

Measurement and Related Philosophical Problems, Joensuu, Finland

1991

Bells Theorem and the Foundations of Modern Physics, Cesena, Italy

1992

Symposia

onthe Foundations

of Modern Physics 1992: The Copenhagen

Interpretation and Wolfgang Pauli, Helsinki, Finland

1993

International Symposium

on Fundamental

Problems in Quantum Physics,

Oviedo, Spain

1994

Fundamental Problems inQuantum Theory, Baltimore, USA

1995

The Dilemma of Einstein, Podolsky and Rosen, 60 Years Later, Haifa, Israel

1996

2nd International Symposium

on Fundamental

Problems in Quantum Physics,

Oviedo, Spain

on Conceptual

1997

Sixth UK Conference

1998

Mysteries, Puzzles, and Paradoxes inQuantum Mechanics, Garda Lake, Italy

1999

2nd Workshop

2000

NATO Advanced Research Workshop

2001

Quantum Theory: Reconsideration of Foundations, V axj

and Mathematical Foundations of

Modern Physics, Hull, England

on Fundamental

Problems inQuantum Theory, Baltimore, USA

on Decoherence

and its Implications

inQuantum Computation and Information Transfer, Mykonos, Greece

o,Sweden

wrong

But there has to be something

this! If
light,

any

with

of these priests had truly shown the

there

simply

year-after-year

would

conference.

clear enough: If we

not
The

be

the

verdict

seems

i.e., the set of people who

care about
us as a
meetings
are

might be reading this paperreally

quantum foundations,
community

to ask

happening and find

then it behooves
why

these

a way to put a stop to them.

My view of the problem is this.

the

accusations

Despite

incompleteness,

of

nonsensicality,

sees one
religion making against the other, I
see little to
no difference in any of their canons. They all
irrelevance,

look

equally

quantum

to want
single

detached

practice to

a firm

God

one

and surreality

from

often

the

me. For, though

reality within the

can

they

point

world

of

seems
theoryi.e., a
each

to and declare,

There, that term is what is real inthe universe

even when

are no physicists about none

very hard to get out of the
Platonic realm of pure mathematics to find it.
What
Imean
by
this
deliberately
have

there

worked

provocative
differences

statement
in what

is that in spite of the

the churches

real in quantum theory,

proceed

from the

same

label

to be

they nonetheless

abstract

starting

all

point

the standard textbook accounts of the axioms

of quantum theory.

The Canon for Most of the Quantum Churches:

The Axioms (plain and simple)

every system,

there is a complex Hilbert

space

1.

For

2.

States of the system correspond to projection operators onto H.

3.

Those things that

are observable

H.

somehow correspond to the

eigenprojectors of Hermitian operators.

4.

But what

Isolated systems evolve according to the Schr odinger equation.

nonsense

is this,

is this, and

you must

no one

Carlo Rovelli [11]

be asking. Where else could they start? The main issue

has said it

more

clearly than

Where present-day

quantum-

foundation studies have stagnated inthe stream

of history is not

so unlike

where the physics of

length contraction and time dilation stood before

Einsteins 1905

paper on special

relativity.

The Lorentz transformations

they

do,

rather

have the

name

say, the Einstein

reason: Lorentz had
as early as 1895. Indeed

than,

transforma- tions, for good

published

one

some
say

could

of them

that

most

the

empirical

were in place well

came onto the scene. But that was

predictions of special relativity

before Einstein

of

of little consolation to the pre-Einsteinian physics

community

so

striving

electromagnetic

ether. Precisely because

the only justification for

appeared to be their

a mystery to
particularly, this was a

empirical adequacy, they

mystery

that

More

of

and the luminiferous

the Lorentz transformations

be conquered.

sense

hard to make

phenomena

remained

heaping

ad

further

hoc

structure onto could not possibly

(mathematical)
solve.
2

Or add to the theory,

Very

might

briefly,

as

be

by

captured
actual

universes
though,

in

Hamiltonian.

these

some

are

universal

state

vector

with

preconceived

tensor-product

Spontaneous

Collapsians

vectorthough

now

for

it

is

For
and

an

the
the

the persuasion,
supplemented

decompositions

respect

is

with

function

upon

struct

the

vector

sometimes

with the terms in various Schmidt

positions

reality

space.

wave

(Depending

two entities

the

Bohmians,

coordinate

universal

be.

of

the state

supplementing

it is the

of

For the

follows.

trajectory

Everettics,

as the case may

a cartoon

ures
again

individual

to

of the

various

) For

the

the

state

systembut

Hamiltonian

dynamics

collapse

mechanism.

reality

is captured

state and
preferred

is supplemented
For

the

with respect to

Hamiltonian

with

of

measures

valued

objective

Historians

an initial quantum

by the addition

positive-operator

an

Consistent

a set

sets of historiesalong

they call them consistent

of

(POVMs)

with

truth-value assignment within each of those sets.

4

To be fair, they do, each in their

minor modifications

own way,

to the meanings of

a few

contribute

words in the

where the effort stops.

axioms. But that isessentially

3
What

was

of

years
was an understanding

being begged for in the

between 1895 and 1905

the origin of that

structuresome

abstract,

simple,

mathematical

crisp

physical

statements with respect to which the necessity of

the

mathematics

Einstein supplied

would
that

be

indisputable.

and became

one

of the

greatest physicists of all time. He reduced the

mysterious

structure

transformations

to

of

two

the
simple

expressible in common language:

Lorentz

statements

space is
of its source, 2)physics

1)the speed of light inempty

independent of the speed

appear

should

the

same

inall inertial reference

frames.
The deep significance

problem

should

up

anyone

to

overpoweringly

of this for the quantum

stand

and

who

speak

admires

these

principles.

move effectively stopped all further

on the origins of the Lorentz transforma-

Einsteins
debate

tions. Outside of the time of the Nazi regime in

Germany [12], I
suspect there have been less than

a handful

of conferences devoted to interpreting

them. Most importantly,

plicity

of Einsteins

ready

for the

imagine

that

next

any

with the

principles,

step.

mindeven

supreme

physics

sim-

became

Is it possible

to

Einsteinscould

have made the leap to general relativity directly

from

the

original,

abstract

Lorentz transformations?
only

Indeed,

find

in

one can

A structure

adequate?

empirically

structure

of

the

that

was

Iwould

dream of the wonders

pursuing

the

same

strategy

say no.
we will
of

simplification for the quantum foundations.

we are:

Symbolically, where

Where

Speed of light

vt

p1

v 2 /c 2

t vx/c

p1

we need to be:

is constant.

/c

Physics is the same

inallinertial frames.

sense of the
more structure,
more defini-tions, more science-fiction imagery on
top of them, but to throw them away wholesale
The

task

quantum

and start
asking

is not

axioms

to make

heaping

afresh. We should

ourselves:

principles

might

mathematical
should

by

From

we

derive

structure?

be crisp; they

be relentless

what

deep
this

Those

should

be

in

physical
exquisite
principles

compelling.

They should stir the soul. When I

was injunior

high school, I
sat down with Martin
book Relativity

Gardners

for the Million [13] and

came

away

an understanding of the subject that

me today: The concepts were strange,
but they were clear enough that I
could get a
grasp on them knowing little more mathematics
than simple arithmetic. One should expect no
less for a proper foundation to quantum theory.
Until we can explain quantum theorys essence
to a junior-high-school or high-school student and
have them walk away with a deep, lasting
memory, we will have not understood a thing
with

sustains

about the quantum foundations.

So, throw

pro-ceed?

the existing

away

mechanics

I
myself

contemplate
techniques,

deep
and

axioms

of quantum

and start afresh! But how to

see no
and

alternative

hard

the

the implications

information theory. The

reason

to

but

tasks,

the

of quantum

is simple, and I

think inescapable. Quantum mechanics has always

been

about

information.

It is just

that

the

physics community has somehow forgotten this.

Quantum Mechanics:

The Axioms and Our Imperative!

States correspond to density

operators

an information
reason if possible!

theoretic

an information

theoretic

Give

over a Hilbert space

H.

Measurements correspond to positive

operator-valued
{Ed

measures

Give

reason if possible!

H is a complex vector

not

(POVMs)

}on H.

space,

a real vector space, not a

Give

an information

theoretic

reason if possible!

quaternionic module.

Systems combine according to the tensor

product of their separate vector

spaces,

HAB

= HA

HB

an information
reason if possible!

theoretic

an information
reason if possible!

theoretic

an information

theoretic

Give

Between measurements, states evolve

according to trace-preserving completely
positive linear

By

maps.

way of measurement, states

Give

evolve

(up to normalization) via outcomedependent completely positive linear

Give

maps.

reason if possible!

Probabilities for the outcomes

of a measurement

obey the Born rule

for POVMs tr(Ed ).

an information
reason if possible!

Give

theoretic

The distillate that remainsthe

piece of quantum theory with no information

theoretic significancewill be our first unadorned glimpse of quantum reality.

Far from being the end of the journey, placing this conception of nature inopen
view will be the start of a great adventure.

This, I
see

as

pursue

with

quantum

system represents

independent

a collection

of

we

should

consistency:

The

the line of attack

relentless

something

real and

us; the quantum state represents

of subjective degrees of belief about

something to do with that system (even if only

in connection

with

our

experimental kicks to it).

The structure called quantum mechanics is about

the interplay

of these two thingsthe

and the objective.

separate
5

the

But physicists

forever

trying

to

The task

subjective

before

us

is to

wheat
are, at bottom, a naive
come to terms

breed,

with the world

out

there by
methods
in

which, however imaginative

essence

the

same

and refined, involve

element of contact

as a well-placed

kick. B.S.DeWitt and R.N.Graham[14]

5
the

from

chaff.

represents
much

subjective
its

of

the

If

same

support

character?

state

then

information,

mathematical

might be of that

quantum

how

structure

Some of it,

maybe most of it, but surely not all of it.

Our foremost

and

an

every

task should be to

go to

each

axiom of quantum theory and give it

information

Only when

theoretic

we are

justification

finished

terms (or combinations

picking

if

we can.

off all the

can
will we

of terms) that

be

as subjective information
be
a position to make real progress in quantum
foundations.
The
raw distillate left
behindminuscule though it may be with respect
to the full-blown theorywill be our first glimpse
of what quantum mechanics is trying to tell us
interpreted
in

about nature itself.

Let

about

me try to

give

this by making

a better way to think

use of Einstein again

What might have been his greatest achievement

in building general relativity?

was

in his recognizing

I
would

say

it

that the gravitational

field

one

coordinate

feels in an accelerating

a
case

elevator is

effect. That is, the field inthat

is something induced purely with respect to the

description

program

an

of

observer.

In this light, the

of trying to develop general relativity

boiled down to recognizing all the things within

gravitational
should

be

and
viewed

coordinate choices.
things that

are

the observer-free

come

about

motional

additional

numerically

situationi.e.,

purely

by

that

[15] to

those things that

bringing

agent, coordinate

(scientific

phenomena

as consequences of our
It was in identifying all the

the

observer

system, etc.) back

into the picture.

This
weeding

was a true
out

interpreted

all

as

behind finally

Riemannian

breakthrough.

the

becomes

manifold

can

be
left

that

clear to sight: It is the

we

call

spacetimea

mathematical object, the study of which

hope will tell

us

in

effects, the fruit

things

coordinate

For

something

one can

about nature itself,

not merely about the observer innature.

The dream I
see for quantum mechanics

is

just this. Weed out all the terms that have to

do

with

gambling

commitments,

information,

knowledge, and belief, and what is left behind

will play the role of Einsteins manifold. That is

our goal. When we find it, it may be little

more than a minuscule part of quantum theory.
But being a clear window into nature, we may
start to see sights through it we could hardly
imagine before.

Summary

I
say to the House

as

I
said to ministers who have joined

this government, I
have nothing to offer but blood, toil,

tears, and sweat. We have before

grievous kind. We have before

us an ordeal

of the most

us many, many

months of

struggle and suffering. You ask, what is our policy? I

say
it is to wage

war. War

with all our might and with all the

strength God has given

can answer

us. You

ask, what is

our

aim? I

inone word. It isvictory.

Winston Churchill, 1940, abridged

This paper is about taking the imperative in

the Introduction seriously, though it contributes
only a small amount to the labor it asks. Just
as in the founding of quantum mechanics, this
is

I
should point out to the reader that in opposition to

the picture

of general relativity,

coordinate

systemi.e.,

observerchanges

tells

us

what

nothing

kind

more

that

the

there), and spacetime

matter

is

there).

necessary part
things

once

mechanics

same

pure
a

spacetime, rather

(when

how

to

system.

matter

move

is

(when

agents,

scientific

Observers,

world

will be

coordinate

curve

matter

tells

[16]

to

will pick

the observer

it off with
how

(it only

manifold

for the quantum

matter into

gridding

tells spacetime

Matter

the

the

the

the observer

reintroducing

like introducing

than simply

about

of sensations

up), I
do not suspect
Here I
suspect

where reintroducing

reintroducing

of reality? No. But do they tend to change

they

are on

the

scene?

can tell us something

Yes. If quantum

deep about nature, I

truly

think it is this.

not something

that will spring

lone mind inthe shelter of

It is

task for

productive

points

community

Philosophy

philosophers,

college.

The

say

to be left to the

John Archibald Wheeler

I
am apt to

quantum

is nothing if not that.

is too important

with diverse but

view.

of

community

information

Likewise,

forth from

a medieval

once

said.

for the quantum

foundations.
The structure of the remainder
is

as

In Section

follows.

paper

Why

Ireiterate

Information?,

of the

the

cleanest

argument I
know of that the quantum state is
solely

an

expression

informationthe

quantum system. It has

and of itself.

considering

subjective

of

one

information

no

The argument

the

has about

objective

reality in

is then refined by

phenomenon

quantum

of

teleportation[23].

In

Section

What?,

I
tackle

head-on.

The

consequences

of

into nature.

quantum

About

Information
that

very

question

answer is the
our experimental

Once freed

measurement

[24]

potential
interventions

from the notion that

ought

to

be

about

some preex-isting property [25]

one finds no particular reason

revealing traces of
(or beable [26]),

to take the standard account of measurement (in

terms of complete sets of orthogonal projection

operators)
information

as a

basic notion.

theory,

utility of generalized

Indeed quantum

with its emphasis

measurements

or

on

the

positive

measures

operator-valued

suggests

one

[27]

(POVMs)

as

should take those entities

basic notion instead.

The productivity

the

of this

point of view is demonstrated by the enticingly

simple Gleason-like

probability

[28] and, independently,

collaborators [29]

of the quantum

derivation

rule recently

found by Paul Busch

by Joseph Renes

Contrary to Gleasons original

theo-rem [30], this theorem works just

Hilbert

spaces over

the

a strengthened argument

noncontextuality

theorem.

In

International

start the

des Poids

assumption

4.2,

Section

process

as well for
even for

and

the field of rational numbers.

In Section 4.1, I
give
for

spaces,

Hilbert

et Mesures

of defining

from the Bayesian point of

in this

Bureau

Le

two-dimensional

and

what

Paris,
it

viewto

means
accept

as a theory. This leads to

a fiducial or standard
quantum measurement
for defining the very
meaning of a quantum state.
quantum mechanics
the

notion

of fixing

In Section 5 Wither
ask whether entanglement

as

far

as quantum

Entanglement?,

is all it is touted to be

foundations

are

concerned.

That is, is entanglement

said,

the

mechanics,

combat

one

the

departure

tensor-product

trait

that

from classical

this, Igive

as

really

characteristic

Schr odinger

quantum

of

its

enforces

entire

lines of thought?

simple

derivation

rule for combining

To

of the

spaces

Hilbert

of individual systems which takes the structure

as its starting
use of

of localized quantum measurements

point. In particular, the derivation makes

considerations

Gleason-like
7

If

you want to

in

know what this

the

means,

ask

me over a

beer sometime.
8

There

have

been

other

soundings

and computation

theory

deep

the foundations

of quantum

about

the

of

can tell us

information

idea

that

something

mechanics.

See

Refs. [17], [18], [19], and inparticular Ref. [20].

9

In

the

previous

quant-ph/0106166,
information

emphasize
probability
nothing

so

and

states

ascribes

than

the

to

of

paper,

quantum

Bayesian

are

states

and

did

not

idea

that

the

phenomenon

to that

Bayesian, probabilities

this

knowledge

gambling

willing to make with regard

radical

of

called

much the radical

one

more

version

I
variously

amounts

commitments

to

one

is

phenomenon. To the
subjective

all the

way

to the bone. Inthis

to turn

my

somewhat

paper, I
start

earlier de-emphasis

dangerous

the long

around

process

of trying

(even though it is

to attempt this in a manuscript

is little

more

paper)

In particular, because of the objective overtones

the

word

knowledge

than

a modification

i.e.,

knowledge

is either right

from the term

as

The conception

much

as

or wrong

no

such thing

beliefs

call

or

mathematical

gambling

quantum

information

more

opinions,

already

commitments.

states

beliefs,

piece

sense

as a

right

paper

is specifically

Thus I
now

states

than is becoming

a set

and gambling

commitments.

of

variously

of

belief,

common

community),

is

and true

for capturing

well the number

of

version.

of this

information

understand

of

try to steer clear

(though, by this I
mean information

subjective

quantum

particular

a quantum state

symbol

completed

in the present

possible

quantum state. In all cases,

that

working in the background

that there is simply

and only

of

an already

that

in a

in the

judgments,

Believe

me,

of jaws that

will

drop from the adoption of this terminology. However, if the

reader finds that this gives him
stomachor

a sense

fears that I
will become

crystal-toting

New

medicine[22]I

to be absolutely

Age

of butterflies

practitioner

in the

solipsist [21]
of

or a

homeopathic

hope he will keep inmind that this attempt

frank about the subjectivity

of

some

of

the terms in quantum

theory is part of

delimit the terms that

fruitful

a larger program to

can be interpreted as objective

in a

way.

presence

of classical

notion

communication.

structure

tensor-product

of entanglement

With the

established,

follows

shows how entanglement,

the

very

in step. This

just like the standard

rule, is secondary to the structure of

probability

quantum measurements. Moreover, locality

is

built in at the outset; there is simply nothing

mysterious and nonlocal about entanglement.
In Section 6 Whither

ask

one

why

Bayes Rule?,

expect

the

quantum state assignments

updating

completion of

it actually

for

upon

the

a measurement to take the form

way, I
give a simple
ones

that

on average

measurement

that

does

a tautology!)

proof technique

always

information

for any quantum mechanical

information. (Despite the

this is not

rule

does. Along the

derivation
increases

should

used for

not

itself

appearance

discard

otherwise,

Most importantly, the

showing the theorem

indicates

an

quantum

collapse

probability

strong analogy

extremely

and Bayes

overall

unitary

of ones final proba-bilistic beliefs

readjustment

the

readjustment

initial

state

takes

the

of

into

as

system

for the

description

an

Up to

theory:

between

rule in classical

interactionquantum

account

ones

as

well

ones

measurement
collapse

is

precisely

Bayesian conditional- ization. This in turn gives

more

impetus

sections.

for the assumptions

theorems

Gleason-like

the

behind

previous

6.1, Accepting

In Section

Mechanics,

of

I
complete

the

process

the

two

Quantum

started

in

Section 4.2 and describe quantum measurement

inBayesian terms: An everyday

any

I-know-not-what

that

mea-surement
to an

is

leads

application of Bayes rule with respect to ones

belief

about

the

outcome

potential

of

the

standard quantum measurement.

In

Section

Information?,

a quantum state

7,

describing

Else

is

I
argue that, to the extent that

a subjective quantity, so must

of a state-change rule d
what
happens
to an initial

is

be the assignment
for

What

quantum

state

upon

the

completion

some

measurementgenerally

of

POVMwhose

outcome is d. In fact, the levels of subjectivity

for the state and the state-change

precisely
draw

an

the

same

rule must be
sake. To

for consistencys

analogy to Bayesian probability

theory,

the initial state plays the role of an a priori

probability

distribution

P(h)

hypothesis, the final state d

posterior

probability

distribution

the state-change

rule d

the
andstatistical
the

model

some

for

plays the role of

P(h|d),

and

plays the role of

P(d| h) enacting

the

transition

P(h) P(h|d). To the extent that all

Bayesian

probabilities

are

subjectiveeven

the

probabilities
P(d|h)

of

a statistical

modelso is the mapping

Specializing to the

is gathered,

one

completely positive
time-evolution

case

that

no information

finds that the trace-preserving

are

maps

that describe quantum

themselves nothing

more

than

subjective judgments.

In Section 8 Intermission,
breather to sum

up what

I
give

a slight

has been trashed and

we are headed.

where
In

Unknown

I
tackle

very

these

Section

States?,

words.

use

ubiquitous

Quantum

the conundrum
Despite

posed

the

in the quantum

information

can an unknown state

literature, what

by

phrases

be?

quantum statefrom the present point of view,

explicitly someones

informationmust

someone, if it exists at
hand, for many an application

known by
other

it

information,

imagine

that

background

is

there

be

quite

is always

describing

grounding

in quantum

contrived

someone

to

in the

system

being

and that what

we are

the

or manipulated,

measured
doing

would

always be

all. On the

the

phenomenon

with

respect to his state of belief. The solution, at

least in the
[31]

case

is found

version

of

of quantum-state

through

de

Finettis

unknown probabilities.

Refs. [32] and

interesting

classic

mechanical

theorem

on

This reports work from

[33]. Maybe

one

of the most

things about the theorem is that it

fails for Hilbert

numbers,

tomography

a quantum

spaces

suggesting

over

that

the field of real

perhaps

the

whole

discipline of quantum information

might not be

well defined inthat imaginary world.

Finally, in Section 10 The
the

I
flirt

Quantum,

tantalizing

Oyster

with

the

and

most

question of all: Why the quantum?

no answer here, but I

do not discount
we are on the brink of finding one. In this
regard no platform seems firmer for the leap
than the very existence of quantum cryptography

There is
that

and quantum computing. The world is sensitive

to our touch.
8

a kind of Zing! 10 that makes it fly off

ways that were not imaginable classically.

It has
in

The whole structure of quantum mechanicsit

speculatedmay

optimal

method

be

nothing

more

of reasoning

than

is

the

and processing

a fundamental
a concrete proposal

information inthe light of such

(wonderful)
for

Zing!,

sensitivity.

potential

I
consider

traditionally

As

mathematical

the integer

ascribed to

expression

parameter

a quantum system

of

by

way

of its Hilbert-space

Why Information?

Realists

can

dimension.

be tough customers

indeedbut

there

is

no

reason to be afraid

of them.

Paul Feyerabend, 1992

was the master of clear thought; I

have expressed my opinion of this with respect
to both special and general relativity. But I
can
go further. I
would say he possessed the same
great
penetrating
power when it came to
analyzing the quantum. For even there, he was
Einstein

immaculately

clear and concise inhis expression.

In particular, he
absolutely

was

the first

unambiguous

person to say

in

terms why the quantum

as information (or, to say

as a representation of ones
gambling commitments, credible or

state should be viewed

the

same

thing,

beliefs and

otherwise).
His

argument

quantum-state

was

assignment

that
a
a system can

simply

for

be

go one way or

forced to

a part

with

of the world that should have

causal connection

The paradigm
known

paper

with

course

the Einstein,

a long

the

one

Podolsky,

but simpler versions

thought had

no

the system of interest.

here is of

through

[34]

the other by interacting

pre-history

well

Rosen

of the train of

with Einstein [35]

himself.
The best

was

in

essence

this.

Take two

spatially separated systems A and B prepared

in

some

entangled

quantum

state

AB

i. By

of one or another
on system A alone, one can
her
immediately
of two
write down a new state for system
B. Either the state will be drawn from one set
Bi
of states {| the
i}state
or another
wil l
be{|
drawn
i}, depending
from one
i
upon which observable is measured. 11 The key

the measurement

performing
of two

observables

point is that it does not matter how distant the

two systems

are

from each other, what sort of

medium they might be immersed in, or

the

other

concluded

fine

that

details

of the

whatever

world.

these

any

of

Einstein

things

called

quantum states be, they cannot be real states of

affairs

for system B alone. For, whatever

the

real, objective state of affairs at B is, it should

upon the measurements one can

ona causally unconnected system A.
Thus one must take it seriously that the new
state (either a | Bi i or a | i i) represents
information
about system B. In making a
measurement on A, one learns something about
not depend
make

B, but that is where

change

more

cannot

the story ends. The state

be construed

to be something

physical than that. More particularly,

than

a reflection

ones

initial

of

some

tricky combination
and

information

the

gained through the measurement.

the language

cannot

be

of Einstein,

complete

the

as more

final state itself for B cannot be viewed

of

knowledge

Expressed

the quantum
description

in

state

of

the

quantum system.
Here is the

way

Einstein put it to Michele

Besso ina 1952 letter[37]:

10

Dash,

11

verve,

Generally

between

vigor, vim, zip, pep, punch, pizzazz!

there

need

be

the two sets of states:

hardly
only

any
that

relation

when

the

states

are

together

weighted

by

their

probabilities,

to form the initial density

B alone. For

precise statement

operator

they

mix

for system

see

of this freedom,

Ref.

[36].

9
What

relation

is there between

quantum state) described by

real deterministic
state

) ?

Does

we

situation (that
the

the state

function and

quantum

call the real

state

characterize

completely (1) or only incompletely (2) a real state?

cannot

One

respond

to

unambiguously

question, because each measurement

represents

uncontrollable

the

(Heisenberg).
something

that

experience,
thetical.

intervention
The

real

is

in

state

immediately

and its appreciation

(Comparable

therefore

to

accessible

always

rests hypo-

to the notion of force

classical mechanics, if one doesnt fix

of motion.) Therefore

system

not

is

suppositions

a priori

can be taken

in

the law

(1) and (2)

inprinciple, both possible. A de-cision in favor of

of them

this

a real

are,
one

an examination and
consequences
obliges us to admit that

only after

confrontation of the admissibility of their

I
reject (1) because it
there is

rigid

system separated

connection

between

from each other in

parts of

space

in

the

an

arbitrary

way

(instantaneous

action at

distance,

, ,, ,

which doesnt diminish when the distance increases).

Here is the demonstration:

A system

with

S12

function

which

12

known, is composed of two systems S1

and S2

is

which

are very far from each other at the instant t.If one
makes a complete measurement
on S1 which can
be done in different ways (according to whether one
measures, for example, the momenta or the
coordinates),
depending
on the result of the
measurement and the function 12 one can determine

by current quantum-theoretical methods, the function

2

of the second

different

measurement

system. This function

to

according

forms,

applied to S1

the

But this is in contradiction

excludes

measurement
S2

at

action

on S1

and therefore

effect

with (1) if

one

assuming

(1)

pass to

the real state

effect

considers

quantum theory

amounts

as

the

a system

described incompletely by the function 12

one

on

method

of

the

complete

real states. One could justify this

(2)

is only

present

being in principle definitive,

to renouncing

the

the supposition

according to which the real state of

If

the

Therefore

on
no

quantum state of S2 described by 2

I
am thus forced to

of

distance.

no

has

can assume

procedure

that

description

re-nunciation

of

if one

assumes

that there is no law for real statesi.e.,

their description
that would

would

mean:

only to what

that

be useless. Otherwise

said,

laws dont apply to things, but

observation

teaches

us

about

them

(The laws that relate to the temporal succession of

this

partial

knowledge

are

however

entirely

deterministic.)

Now,
statistical

accept

Icant
character

conditioned

by

that.

Ithink

of the present

the

choice

of

theory

an

that

the

is simply

incomplete

description.

There are two issues in this letter that are

worth disentangling.
1) Rejecting
the rigid
12that
connection
of all nature
is to say,

admitting

that

systems has

any

the

the conclusion that

a complete

very

meaning

notion

of

separate

at allone is led to

a quantum state cannot

a system. It must

specification of

be

be

information, at least in part. This point should

be placed in contrast to the other well-known

facet

of

Einsteins

unwillingness

as a necessary

thought:

to accept such

an

namely,

2)

an

incompleteness

trait of the physical world.

It is quite important

to recognize that the

first issue does not entail the second. Einstein

had that firmly in mind, but he wanted
His

reason

for going the further

...

step

more.
was, I

think, well justified at the time [38]:

There exists

reason

a simple

psychological

for the fact that this most nearly obvious

interpretation is being shunned.

For if the

statistical quantum theory does not pretend to

describe the individual system (and its

development in time) completely, it appears

unavoidable
12

The rigid connection of all nature,

on the other

hand,

is exactly what the Bohmians and Everettics do embrace,

even glorify.

So, I
suspect these words will fall on deaf

ears

with them. But similarly would they fall on deaf

ears

with the believer who

every event

says

inthe universe and

that God wills each and

no further

explanation is

needed. No point of view should be dismissed out of hand:

the overriding issue is simply which view willlead to the

most

progress,

which view has the potential to close the

debate, which view will give the most

new

phenomena

for the physicist to have fun with?

10
to look elsewhere for

complete description

individual system; in doing

so it would

of the

be clear from

very

the

beginning

description

are not

that

the elements

contained

within

of such

the conceptual

scheme of the statistical quantum theory. With this

one

would admit that, in principle, this scheme could

not serve

as the basis

of theoretical physics.

The last seventeen

seen much in the mean time.

years have given confirmation

after confirmation

that the Bell inequality

several variations

of it)

But the world has

the

are

indeed violated

physical world. The Kochen-Specker

theorems have been meticulously

point

where simple textbook

by

no-go

clarified to the

pictures

drawn of them[39]. Incompleteness, it

here to

(and

stay: The theory prescribes

can be
seems, is
that no

we know about a quantum

systemeven when we have maximal information
13
about it
there will always be a statistical
residue. There will always be questions that we
can ask of a system for which we cannot predict
matter how much

the

outcomes.

In

quantum

information

is simply

[40]

neither

But

Wolfgang Pauli

theory,

not complete

can

it be

maximal

information

completed.

once wrote to Markus

As

Fierz [41],

The

well-known

mechanics

be

removed

by

an

existent

but certainly

cannot

is certainly

fact somehow-somewhere,

quantum

of

incompleteness

(Einstein)

reverting

to

classical

field

physics. Nor, I
would add, will the mystery of
that existent

fact be removed by attempting

to give the quantum state anything resembling

an

ontological status.
The

complete

disconnectedness

the

of

quantum-state change rule from anything to do

with

spacetime

something

in the

intolerable.

The

is

telling

quantum

state

Subjec-tive, incomplete

information.
Put

considerations

deep:

right

It is

mindset,

a statement

this

us
is

information.
is

about

so

not

our

world.

There is something about the world that keeps

us

from

ever

be captured

getting

more

through

infor-mation than

can

the formal structure

of

quantum mechanics. Einstein had wanted

look furtherto

find out how

us to

the incomplete

information could be completedbut

perhaps the

can it not be completed?

Indeed Ithink this is one of the deepest
questions we can ask and still hope to answer.
real question is,Why

But

things

first

question for

one

first.

anyone

more immediate
come this farand

The

who has

to be answered

that deserves

forthrightis

what is this information

to be
symbolized
answe re dby
forth
a righ
|i

actually about? I
have hinted that I
would not
dare

say

variable (as the Bohmian

our

some

that it is about

kind of hidden

might)

or even

about

place within the universal wavefunction

(as

the Everettic might).

Perhaps the best

way to

build

up to an

answer is to be true to the theme of this

paper: quantum foundations in the light of
quantum
information.
Let
us forage the
phenomena of quantum information to see if we
might first refine Einsteins
look

no

quantum

further

argument. One need

to the

than

teleportation

[23]

quantum-state assignment for

to go

one way or the

another

part

of

phenomenon

Not

only

a system

can a

be forced

other by interacting with

the

world

of

no

causal

significance,

but, for the cost of two bits,

can

that

make

anything

one wants

of

quantum
it to be.

state

one

assignment

Such

an experiment starts out with Alice and

a maximally entangled pair of qubits

Bob sharing

inthe state r12

|

AB

(1)

(|0i|0i

Bob then

goes to any

|1i|1i)

place in the universe he

laboratory
prepares
any state |i that she
ultimately wants qubit
to impart
wit hany
ontostate
Bobs
| i
system.
t hat
She performs a Bell-basis measurement on the
two qubits in her possession. In the same vein as
wishes

Alice

another

qubit

thought

Einsteins

13

in her

with

As should be clear from all

longer

now,

entirely

pleased

refer to

for instance,

maximally

rigid

my

with this

gambling

warnings,

terminology.

I
am

no

I
would

a pure quantum state as a

commitment

or some

such

thing. See Ref. [2], pages 4950 and 5354. However, after
trying to reconstruct

conformity

with my

a more accurate
the section

this paragraph

new

representation

even more

several times to be in

terminology, I
finally decided that
would

than this footnote!

break

the flow of

11
experiment,

system immediately

Bobs

the character of

one

of the states

takes

on

x |i,

|i,

y
on
|i,
the
or
characte
z |i. r
But
of on
that
e ofisthonly
e stat
insofar
es |i,
asxAlice
|i
is
y |i,
concerned.

14

Since there is

no

(reasonable)

causal connection between Alice and Bob, it must

be that these states represent the possibilities

for

Alices updated beliefs about Bobs system.

If

now

Alice broadcasts

measurement

may

complete

protocol by performing

the four Pauli rotations

conditioning

receives. The result,

is that Bobs

,,

to the world, Bob

the teleportation

system,

the result of her

it

as

(I, x

one

of

z ) on his

on the information he
as Alice is concerned,

far

system finally resides predictably in

the state
system
final
|i.ly resi des predi ctably in the sta te |i
1516

How

can Alice

convince herself that such is the

case? Well, if Bob is willing to reveal his location,

she just need walk to his site and perform the

|ih| vs. I
|ih|.
a YES with probability one
gone well in carrying out the

YES-NO measurement:

The outcome will be

for her if all has

a measurement on a

protocol. Thus, for the cost of

system and two bits worth

causally disconnected

of causal
i.e.,

action

one

on

the system of actual interest

of the four Pauli rotationsAlice

sharpen her

can

to complete certainty

predictability

for any YES-NO observable she wishes.

Roger

has

a state

property

argues

Penrose

in his

book

The

a system
to be some

New Mind [42] that when

Emperors

|i

there

ought

in the system (in and of itself) that

corresponds

to its |iness.

For how else could

the
corr system
espon ds
betoprepared to reveal a YES in the
case that Alice actually checks it? Asking this
rhetorical question with a sufficient amount of
command is enough to make many a would-be
informationist weak in knees. But there is a
crucial oversight

we

have already

implicit

in its confidence,

caught it in action. If Alice

fails to reveal her information

the world, there is

the qubits

no one

ultimate

More importantly,

and

to anyone else in

else who

revelation

can predict

with certainty.

there is nothing in quantum

mechanics that gives the qubit the

power to stand

up

and

say

YES all by itself: If Alice does not

take the time to walk

it, there is

with
17

consequence
4

revelation.

in Alices

confidence

interact

no

over to it and
There

mind

it, she

interact with

is only the

that, should

could

predict

she

the

of that interaction.

Information About What?

I
think
student

that

gets

the

sickliest

it, is that

notion

of

it is the

physics,
science

even
of

if

masses,

think that the healthiest

molecules, and the ether. And I

notion,

even

if

student

physics is the science

does not wholly

of the

ways

get it, is that

of taking hold of bodies

and pushing them!

W.S.Franklin, 1903

There are great rewards in being a new

parent. Not least of all is the opportunity to
have a close-up look at a mind in formation. Last
year, I
watched my two-year old daughter learn
things at a fantastic rate, and though there were

untold lessons
for her, there were also a
sprinkling for me. For instance, I
started to see
her come to grips with the idea that there is a

world

independent

14

As

far

as

Bob

is

concerned,

nothing

whatsoever

changes about the system in his possession: It started in

the
15

mixed state

completely

As

far

as

Bob

12

way.

I
and remains that

is concerned,

nothing

whatsoever

changes about the system in his possession: It started in

mixed state

the completely
16

The

repetition

typographical
17

I
adopt

Savages

the

when

I
and remains that

these

this

the

and

to be

terminology

person,

the

decisions,

acts

measurement.

in

world,

via

Inthis

Savage

the

paper,

to L. J.

similar

2, where

the

consequence

to the person,
he

way.

not

is

footnotes

error.

decision theory. A
happen

12

book, Ref. [43], Chapter

terms

acts,

in

he discusses

consequences,

context

of

is anything
writes,

capacity

of

where

rational

that

we

may

add

quantum

I
call what Savage calls the

person the agent, scientific agent,

or observer

instead.

12
of her desires. What struck

me was

the contrast

between that and the gain of confidence

saw

grow

in her

that

existence she could control.

hand. She pushes

on the

I
also

are aspects
The two go hand

there

of

in

world, and sometimes it

gives in a way that she has learned to predict, and

sometimes

it pushes back in a

way

she has not

(and

foreseen

could

may never

manipulate

be able to). If she

to the complete

the world

desires

of her willI

would

be little

became

difference

convincedthere

between

wake

and

dream.

The main point is that she learns from her

forays into the world. In my cynical moments,

I find myself thinking, How

shes

learned

anything

can she

think that

at all? She has

no

theory of measurement.

She leaves measurement

completely

How

undefined.

can

a
a theory

she have

stake to knowledge if she does not have

of how she learns?

Hideo Mabuchi

measurement

once

told

me, The quantum

a set of people.

problem refers to

And though that is a bit harsh, maybe it also

contains

a bit

of the truth.

community making

use

With the physics

of theories that tend to

years, we are apt to

are built
from the top down, not from the bottom up. The
experiment is the basis of all which we try to
describe with science. But an experiment is an
active intervention into the course of nature on
last between 100 and 300

forget that scientific views of the world

part

the

this

or

experimenter;

that experiment

to

see

it

of nature from afar [44]

contemplation

up

the

of

is

not

We set

how nature

reacts. It is the conjunction of myriads of such

interventions
record into

and their

our data

consequences

books.

that

we

18

that we have learned

we can distill from the
data a compact description of all that was seen
andeven more tellinglywhen we can dream
up further experiments to corroborate that
We tell ourselves

something

new

when

description. This is the minimal requirement

of

science. If, how-ever, from such

can

further

reality

distill

our interventions, then

no bone to pick with
It is the most solid thing we can hope for
independent

so much the better.

reality.
from

example

model

a description we
of a free-standing

of

I
have

theory. Classical physics is the ultimate

in that regard.

description,

but it

can

us a compact
more if we

It gives

give much

want it to.
The important
that there is

no

thing to realize, however, is

logical necessity

world-view always be obtainable.

that such

If the world is

we can never

such that

experimental
prepared

realitya

interventions,

then

our

of

we must

be

for that too. That is where quantum

in its most

theory

identify

realityindependent

free-standing

minimal

and conceptually

seems to stand [46]. It is a

theory whose terms refer predominately to our
interface with the world. It is a theory that
cannot go the extra step that classical physics
did without
writing
songs Icant
simplest dispensation

18

But I
must stress that I
am not

think

that

primitive
of

the

physics

Vienna

an

No

Heisenberg

are by

one got

(in

closer

a quote

he

positivistic

as to

grounded

on a

be

as

the philosophers

interventions

experimenter

to be thoroughly

but

behind.

did. The

Circle

that

that, in a sense, they

years

so

somehow

notion of sense impression

consequences
option

should

and their

records,

theory-laden.

necessity

attributed

one

at least

to the salient

have

no

It is just
theory

point

to Einstein

than

many

after the fact)[45]:

It

is quite

wrong

to try

founding

observable magnitudes alone. Inreality the

happens.

can observe.

very

theory

on

opposite

It is the theory which decides what

You must appreciate that observation

we

is

a very

process.

complicated

The phenomenon

under observation produces certain events inour

measuring

apparatus. As

a result,

further

take place inthe apparatus, which eventually

by complicated
and help

paths

us to fix the effects

produce

processes

and

sense

impressions

inour consciousness.

Along this whole pathfrom

its fixation in our consciousnesswe

tell how nature functions,

the phenomenon

to

must be able to

must know the natural

laws at least inpractical terms, before

we can claim

to have observed anything at all. Only theory, that

is,knowledge of natural laws, enables
the

underlying

impressions.

we

When

us to deduce

phenomena
claim

that

from

we

our
can

sense
observe

something

new, we

ought

really to be saying that, although

we are about to formulate new

agree

not

assume

with

the

natural laws that do

old

ones, we

that the existing lawscovering

from

the

upon them

to

phenomenon

consciousnessfunction

in such

a way

and

hence speak of observation.

nevertheless

the whole path

our
that

we can

rely

13
believe,

with words

that

rhyme

[47]. It

not

observables,
19

dingables.

is

about

We tap

tear and strain to

theory

not

beables,

about

but

bell with

our

about
gentle

touch and listen for its beautiful ring.

are

So what

the world?

ways we can intervene on

are the ways we can push it

the

What

and wait for its unpredictable

usual textbook

are

story is that those things that

measurable

operators.
modern

Or

Xi

(2)

to

set

operators {i } over
that form

to

correspond

perhaps

language,

corresponds

The

reaction?

a complete

of

to
each

say

it

in

observable

orthogonal

complex

Hermitian

Hilbert

more
there

projection

space

HD

resolution of the identity,

=I

The index i
labels the potential outcomes of the

measurement

(or intervention

to slip back into

the

language

possesses

observer

most

promoted

generally

the

probability

P(i)

tr(i )

(3)

density

dictates

that he

outcomes

various

.
.

captured

mixed-state

mechanics

by

way

rule is the only rule that satisfies

of

noncontextuality

outcomes [48]

measuring

two distinct

then

independent

the

PD

or

{i } and
projector

outcome

is

is this. Let

associated

space

it is associated

the statement

complex) Hilbert

contemplates

a single

of which observable

be the set of projectors

(real

one

of

simple

measurement

observables

probability

with. More formally,

a very

for

In particular, if

{i } which happen to share

rule

of Andrew Gleasons amazing 1957

For, it states that the standard

theorem [30]

kind

with

The best justification for this probability

comes

an

When

the information
by

operatorquantum

can expect

above)

HD

and let f :
PD [0,1] be such that

with

for D 3,

Xi

=1

f(i )

(4)

a set

whenever
observable.
exists

f()

of

projectors

The theorem

{i } forms

concludes

that

an

there

density operator such that

tr()

(5)

In fact, in

single blow, Gleasons

derives not only the probability

theorem

rule, but also

the state-space structure for quantum mechanical

states (i.e., that it corresponds to the

convex set

of density operators).

In itself this is

no

that makes the theorem

is the sheer difficulty

Note that

an

no restrictions

is

no

differentiable,

theorem

prove

it [49]

have been placed

assumption

nor that

amazing

required to

the function f beyond the

There

small feat, but the thing

ones

upon

mentioned above.

that

it

need

be

it even need be continuous.

All of that, and linearity

too,

comes

structure of the observablesi.e.,

from the

that they

are

complete

sets of orthogonal

linear vector

space.
one

Nonetheless,

should

projectors

ask:

onto

Does

this

a clearer
vision of where the probability rule comes from?
Astounding feats of mathematics are one thing;
insight into physics is another. The two are often
theorem

really

give

the

physicist

at opposite ends of the spectrum. As fortunes

turn,

can

unifying strand

quantum

foundations

be drawn by viewing

in the light

of quantum

information.
The place to start is to drop the fixation that
the basic set of observables in quantum mechanics

are

complete

sets of orthogonal projectors. In

quantum information theory it has been found

to

be

extremely

convenient

to expand

the

notion of measurement

to also include general

positive

measures

(POVMs) [39,

in place

of the

50]

In other

textbook

19

14

operator-valued
words,

notion

Pronounced ding-ables.

usual

any set

of measurement,

positive-semidefinite

a resolution

(6)

on HD

that forms

of the identity, i.e., that satisfies

|i 0,

h|Ed

{Ed } of

operators

for all |i HD

and
Xd

=I

Ed

(7)

as a measurement. The outcomes of the

are identified with the indices d,
and the probabilities of the outcomes are
computed according to a generalized Born rule,
counts

measurement

P(d)

= tr(Ed )

(8)
The

set

{Ed } is called

POVM,

and

the

Ed are called POVM elements. (In

non-standard language promoted earlier, the
set {Ed } signifies an intervention into nature,

operators
the

while the

individual

represent

Ed

the potential

consequences
nature, while the
of
that
standard measure- ments,

there is no limitation

the number

the index

of values

Moreover, the Ed
is

no

intervention.)

may

requirement

be of
that

any

Unlike

can

on

take.

rank, and there

they

be

mutually

orthogonal.

way

The

measurement
POVM

this expansion
is usually

of the notion of

justified

is that

any

can be represented formally as a standard

on an ancillary system that has

measurement

in the past with the system of actual

interacted

interest. Indeed,

are initially
and A

suppose

the system and ancilla

described by the density operators S

respectively. The conjunction of the two

systems is then described by the initial quantum

state
SA

= S

(9)

Aninteraction between the systems via

some

unitary time evolution leads to a new state

 USA U

SA

(10)
Now, imagine

via

on the total

on the
space

Hilbert

of orthogonal projection operators {I

a set

d }. An outcome
standard

P(d)

measurement

standard

ancilla. It is described

d will be

found,

by

the

Born rule, with probability

= tr

)U

(I

U(S

(11)

The

number

of

outcomes

in this

indirect

notion of measurement

by the

dimensionality

spacein

principle,

seemingly

is limited only

of the ancillas
there

can

be

Hilbert

arbitrarily

many.
As

advertised,

probability

it

turns

formula above

terms of operators

on the

can

out

that

systems Hilbert

alone: This is the origin of the POVM. If

|s iand |ac ibe an
we
system
let |sand
i
andancilla

the

be expressed

in

space
we let

orthonormal

basis

for the

respectively,

then

|s

i|ac

will be

basis for the composite system. Using

the cyclic property of the trace in Eq. (11),

we

get
P(d)

=
(s

Xc

hs |hac |

=
Xc

|s

)U

(I

d )U

i|ac i

hs

!|s

(I
hac |

|ac

A )U

(I

d )U

(12)
Letting trA and trS denote partial traces

over

the

system and ancilla, respectively, it follows that

P(d)
(13)

15
where

= trS (S

Ed

(I

Ed

trA

A )U

(I

( I

d )U

(14)

on the Hilbert space of the

proves half of what is
needed, but it is also straightforward to go in the
reverse directioni.e., to show that for any
POVM {Ed }, one can pick an ancilla and find
operators
any
A
U, and d such that Eq. (14) is
is

an operator

original

acting

system.

This

true.

Putting this all together, there is

which
everything

can

that

be said

measurement theory [50]

a way to

think

measurements

standard

a sense

quantum

about

As became clear above,

about this is that by learning

something about the ancillary system through

measure- ment,

standard
something
Indirect

about

though

one

the system
it

may seem,

in

turn

this

can

technique,

information

that could not have been

sender has only

sending

one

of

learns

of real interest.

powerful

otherwise [51]. A

in

capture

sometimes

be

revealing

re-vealed
very simple example is where a
a single qubit available for the
three potential messages. She

message in one
of three preparations of the system, even though
the system is a two-state system. To recover as
much information as possible, the receiver might
(just intuitively) like to perform a measurement
therefore has

a need to encode

the

with three distinct

outcomes.

were

to

limited

measurement,

If, however,

he

quantum

standard

he would only be able to obtain two

outcomes. Thisperhaps

surprisinglygenerally

degrades his opportunities for recovery.

What I
would like to bring

way

this standard

of justifying

up

the POVM is

the most productive point of view

any

Might

of

the

mysteries

is whether

one can take

of

quantum

mechanics

be alleviated by taking the POVM

notion

basic

POVMs utility

of

measurement?

portend

a larger

Does

as

the

role for it in the

foundations of quantum mechanics?

Standard

Generalized

Measurements

Measurements

{i }

{Ed }

h|i |i 0,|i
Pi

P(i)

=ij

|i 0,|i

Pd

=I

=tr(i

i j

h|Ed

Ed

P(d)

=I

= tr(Ed

I
try to make this point dramatic

lectures

by exhibiting

table above.
list

of

notion

right-hand

of

a transparency

in

my

of the

On the left-hand side there is

various
of

properties

for

the

standard

a quantum measurement. On
side, there is an almost identical

properties

for

the

POVMs.

The

the

list

only

difference between the two columns is that the

right-hand

one

is missing

condition required

of

the orthonormality

standard

The question I
ask the audience

measurement

is this: Does the

addition
make

one extra assumption really

process of measurement
any less

of that

the

mysterious?

Indeed, I
imagine

myself

teaching

quantum mechanics for the first time and taking

a vote

with the

best

audience

of

all, the

students. Which set of postulates for quantum

measurement

would

you prefer?

I
am quite

sure

they would respond with a blank stare. But that

16
is the point! It would make
them, and it should make

no

no

difference

to

to

us

difference

The only issue worth debating is which notion of

measurement

will allow

us to see more

deeply

into quantum mechanics.

Therefore

let

us pose

the

question

that

Gleason did, but with POVMs. In other words,

us suppose that the sum total of ways an

experimenter can intervene on a quantum system
corresponds to the full set of POVMs on its Hilbert
space HD It is the task of the theory to give him
probabilities
for the various consequences of his
let

interventions.

Concerning

those probabilities,

let

us

assume only that

whatever the probability for a given consequence
Ec is, it does not depend upon whether Ec is
associated with the POVM {Ed } or, instead, any
is
}.This
associated
other one with
{ Ed
t
means we can assume there
exists a function
to Gleason)

(in analogy

f :ED [0,1]

(15)
where nE
ED

:
0 h|E|i

1 |i HD

(16)

a POVM,

such that whenever {Ed } forms

=1

f(Ed )
(17)

(Ingeneral,

we will call any function

Xd

f(E) 0

(18)

and

satisfying

= constant

f(Ed )

a frame

function, in analogy to Gleasons

nonnegative frame functions. The set ED is often

called the set of effects
It will come

over

HD .)

as no surprise,

of course, that

Gleason-like theorem must hold for the function

inEq. (15). Namely, it can be shown that there

must exist
f(E)

a density operator

= tr(E)

for which

(19)
This

was

recently shown by Paul Busch [28] and,

independently,

by Joseph Renes and collabora-

tors [29]. What is surprising however is the utter

simplicity of the proof. Let
thing right here and
First,

consider

us

exhibit the whole

now.

the

case

where

HD

and

the

operators
First, con sider
on itth
are
ecase
defined
whe re
only
H Dover the field
(complex) rational numbers.

see

that

f is linear

combinations

of operators

It is

no

with respect

of

problem to

to

positive

,,

thatpositiv
never ecom
go outside
bin ation
ED

For
consider ta
three-element
of operators
hat
nevergo outPOVM
side ED{E1
By assumption f(E1 )PO
+ f(E2
VM)+
{E1
f(E3 )

E2

E3

}.

= 1.
}.However,
By ass

we
mption
can f(
also
E1 group the
POVM to obtain a new
have f(E1
function

= 1. In other

) +f(E3 )

f must

be

additive

E2 )

f(E1 )

f(E2 )

(20)
Similarly for

in this

words, the

respect

to

any two integers m and n,

= mf

f(E)

with

operation:

fine-graining
f(E1

E2

first two elements

POVM, and must therefore

= nf

1
E

(21)
Suppose

n
m

1. Then if we write E

statement becomes:

n
f(G)

= nG, this

(22)

17
Thus

on ED
Thus

we immediately
we imme diately

on One
ED

have

a kind

of limited linearity

h ave a kind of limite d linea rity

might imagine using this property to

the theorem in the following

2-dimensional

way.

Clearly

cap

off

the full

vector
space OD
of
Hermitian
on
HD is
the
setitia
ED n
since
that
v ector
spspanned
ace O D by
of H
erm
operator
set
among other things,
ED
all the projection
on contains,
HD
operators. Thus, we can write any operator E ED as
a linear combination
D

operators

2
D

Xi=1
Xi=1

i Ei

Xi=1

(23)

for

some

fixed operator-basis

f would then give

2
D

f(E)

Xi=1

{Ei

i=1

Linearity of

f(Ei )

Xi=1

(24)

So, if we define

by solving

= f(Ei )

tr(Ei )
(25)

the D

linear equations

we would
f(E)

have
Ei

Xi

tr

Xi

= tr

= tr(E)

i Ei

(26)
and

essentially

normalization
density

be

done.

(Positivity

of f would require to be

operator.)

But

the

problem

and

an actual

is that

in

no guarantee that the

coefficients i can be chosen so that i Ei ED E D
What remains to be shown is that f can be
extended uniquely to a function that is truly linear
on OD This too is rather simple. First,
line artake
on O Dany
positive semi-definite operator E.We can always find
a positive rational numbercan
g such
alw ays
that
fin d
Ea
=pos
gG
itive
and rat
G
ona
EDl
nu
Therefore,
mbe rg s uch
we can
that
simply
E = gG
define
an d
f(E)
G
ED gf(G).
To see that this simply
definition
is unique,
suppose
there
de fine
f(E )
gf(G)
are
To see
twothat
such
thisoperators
defini tio nG1
is uniq
and
ue,su
G2 ppos
(with
e t he
corresponding
are tw o such numbers g1 and g2 ) such that E =
g1G1 = g2 G2 Further suppose g2 g1 Then G2 =
expansion

(23)

there

is

.
.

g1
g2

G1

and,

unextended

by

the

definition

g1f(G1). Furthermore
additivity

homogeneity

of f,

this

we

of

the

original

obtain g2 f(G2 )

extension

retains

the

of the original function. For

suppose

that

neither E nor G, though positive semi-definite,

are

necessarily

.. .
in ED

can

We

find

rational
nec essa

positive

rily
number
in ED
c 1
We
such
c an
thatfind(Ea+
pos
G),itiv E,
e rat
and ional
G are
nu
1c

1c

we

all in
Then,
mber
c ED1
such
that by the rules
obtained,

?1c

?1c

= cf
(27)

Let

space

+ cf

us now

OD

already

(E + G)
?1c

have

= cf

f(E + G)

1c

= f(E) + f(G).

further extend fs domain to the full

This

can

be done by noting

operator H can be written

G of two positive

as

that

the difference H

semi-definite

any

= E

operators. Therefore

define f(H) f(E) f(G), from which it also

,., , ,

follows

that

E1 E2
G2

(as

G1 and

f(G).

suppose

there

= E2

+ G2

extended

previous

in the

we obtain

extension,

immediately.

see

are

four operators

= E1

that

G1

+ G1

this

= E2

Applying f

paragraph)

to this

= f(E2 )+f(G1 ) so

f(E1)+f(G2 )

that f(E1) f(G1 )

new

To

such that H

G2

It follows that E1

equation,

this

f(G)

definition is unique

= f(E2 )

f(G2 ).Finally, with

full linearity

can be checked
as far as the

This completes the proof

(complex) rational number field is concerned: Because

f extends uniquely

to

can indeed go through

(26) without worry.
There

are two

linear functional

on

OD

we

the steps of Eqs. (23) through

things

that

are

significant

about this much of the proof. First, in contrast

to Gleasons
to bar the

original theorem, there is nothing

same

logic from working when D

2. This is quite nice because

community

has

gotten

into

much of the

the

habit

of

thinking that

18
there is nothing particularly quantum

mechanical
because

about

a single

qubit.[52] Indeed,

orthogonal projectors

on H2 can be

mapped onto
orthogonal
antipodes projectors
of the Bloch
on sphere,
H2
it is
known that the measurement-outcome statistics

for any standard measurement

mocked-up through

can be

noncontextual

hidden-variable theory. What this result shows

is that that simply is not the case when

considers the full set of POVMs

as ones

one

potential measurements.

20

The other important thing is that the theorem

works for Hilbert

spaces over the rational

number

field: one does not need to invoke the full power of

the continuum. This contrasts with the surprising

result of Meyer[54] that the standard Gleason

theorem fails insuch a setting. The present

theorem hints at a kind of resiliency to the

structure of quantum mechanics that falls

through the mesh of the standard Gleason result:
The probability rule for POVMs does not
actually depend

so much upon the detailed

workings of the number field.

The final step of the proof, indeed, is to
show that nothing

goes awry

when

we go the

extra step of reinstating the continuum.

we need to show that

on the set ED of
complex operators) is a continuous
ED
function. This comes about in a simple way
In other words,

function f (now defined

from
fs additivity. Suppose for two positive
semi-definite

operators E and G that E G

the

(i.e., GE is positive semi-definite). Then

trivially there exists

a positive

semi-definite

operator Hsuch that E + H

= G and

through which the additivity of f gives f(E)

c be an irrational number, and

sequence and bn a
decreasing sequence of rational numbers
that both converge to c.It follows for any
f(G). Let

let

an

be an increasing

,
.

positive semi-definite operator E,that

f(an E) f(cE) f(bn E)

(28)

which implies

an f(E) f(cE) bn f(E)

(29)

Since liman f(E) and limbn f(E) are identical, by

the pinching theorem of elementary

calculus,

they must equal f(cE). This establishes that

can consistently
f(cE)

(30)

= cf(E)

define

we

Reworking

the extensions

of f in the last

inset (but with this enlarged notion of homo-

one completes the proof in a

manner.
Of course we are not getting something from
nothing. The reason the present derivation is so
easy in contrast to the standard proof is that
mathematically the assumption of POVMs as the
geneity),

straightforward

basic

notion

of

measurement

is significantly

stronger than the usual assumption. Physically,

though, Iwould

add

extra

measurement

say

it is just the opposite. Why

to

restrictions
when

they

from basic assumption

the

notion

to practical

of

the route

only make

usage more

circuitous than need be?

Still, no assumption should be left unanalyzed

a chance of bearing fruit. Indeed, one

what is so very compelling about the

if it stands

can

ask

noncontextuality
assignments)

property

that

both

theorem and the present

(of

intervention

into the

original

use of
as a kind of
world, one might

version make

Given the picture of measurement

invasive

probability

Gleasons

expect the

very

opposite.

why measurement

upon

context

the whole

interaction.

One is left wondering

do not depend

probabilities
of

the

measurement

Why is P(d) not of the

mea
form
surem
f(d,
en

})? Is there

any

good

reason

for this kind of

assumption?

20

In fact,

one

need not consider

in order to derive

of

Kochen

noncolorability

and Specker

3-outcome

only

for

of

POVMs

the full set of POVMs

result along the lines

single
the

qubit. Considering

so-called

trine

or

Mercedes- Benz type already does the trick.[53]

19
4.1

Noncontextuality

In point of fact, there is: For,

the

noncontextuality

assignments

that

for measurement

basic

than

measurements

even

the

(i.e.,

probability

outcomes is

particular

that

Noncontextuality bears

of

one can argue

they

more

structure
be

of

POVMs).

more on how we identify

what we are measuring than anything to do with

a measurements invasiveness upon nature.

Here is

a way to see

that. [55] Forget about

quantum mechanics for the moment and consider

a more

general

worldone

details of quantum

notions

of

systems,

consequences,
notion of a

that, skipping

mechanics,

the

still retains

machines,

actions,

the

and

and, most essentially, retains the

scientific

agent

performing

those

consequences.
imagine acting on

actions and taking note of those

a system

Take
with

we

one

S and

might colloquially

call measurement

devices

, ,.

a theory like quantum

case of machine M, let us label
the possible consequences of that action {m1 m2
.}.(Or if you want to think of them in the mold
if

we

it

of two machines, M and Nthings that

had

mechanics.

the aid of

For the

of quantum mechanics,

outcomes.)

For the

call them measurement

case

of machine

N, let

us

label them {n1,n2 ,...}.

If

one

takes

Bayesian point of view about

probability, then nothing

can stop

the agents in

this world from using all the information available

to

them

consequences

to

ascribe

probabilities

to

the

of those two potential actions. Thus,

for an agent who

probability

cares to take note, there are two

distributions,

pM (mk ) and pN (nk ),

lying around.

These

probability

stand for his

subjective

judgments

distributions

about what

will obtain if he acts with either of the two

machines.

a
more. Let us suppose
the labels mk and nk are, at the very least, to
be identified with elements in some master set
F
that is, that there is some kind of
mas
connective
ter set F
glue
for comparing the operation of
one machine to another. This set may even be a
set with further structure, like a vector space or
This is well and good, but it is hardly

physical theory. We need

something, but that is beside the point. What is

of first

concern

is under what conditions

agent identify two particular

with the

same

disparate

in

element

labels mi

will

and nj

F in the master

appearance,

an

construction,

set
and

history though the two machines Mand Nmay

be. Perhaps
Lucent

one

machine

Technologies

was

while

manufactured
the

other

by

was

manufactured by IBM Corporation.

There is really only

one

tool available for

the

purpose,

assignments
pM

(mi )

pM

6=

namely

the

probability

(mi )and pN (nj ).If

pN (nj )

(31)

then surely he would not imagine identifying mi

and nj

with the

same

element F F. If, on the

other hand, he finds

pM

(mi )

pN (nj )

(32)

regardless of his initial beliefs about S, then

we

might think there is some warrant for it.

That is the whole story of noncontextuality.

more than: The consequences (mi

and nj ) of our disparate actions (M and N)
should be labeled the same when we would bet
the same on them in all possible circumstances
(i.e., regardless of our initial knowledge of S).
To put this maybe a bit more baldly, the label by
which we identify a measurement outcome is a
subjective judgment just like a probability, and
just like a quantum state.
By this point of view, noncontextuality is a
It is nothing

tautologyit

is built in from the start. Asking

why

we

have it is

do have

one

a set

a master
may tell us
of

that

we

we

make

set

over

something

Why should the mi s be drawn

of effects ED ?Not all choices of the

master set
settled

choice

Asking

about physics.
from

of time. Where

freedom is in asking why

particular

another.

a waste

on

are

equally interesting

noncontextuality
21

assignments.

once we

have

for the probability

But quantum mechanics, of course,

is particularly interesting!

21

See

Ref.

[56],

pp.

8688,

and

Ref.

[57]

for

some

examples inthat regard.

20
Le Bureau International des Poids

et Mesures

There
important,
the basic

a Paris

4.2

is

still

one

advantage
notion

further,

particularly

to thinking of POVMs

of measurement

as

in quantum

an appropriately chosen
single POVM one can stop thinking of the
quantum state as a linear operator altogether,
mechanics.

For

with

and instead start thinking

judgment

outcomes

of

the

International

Measures.

Our

22

(potential)

measurement.

device right next to the

meter in

outcomes
operator.

for

Bureau

of

Such

on

hinges
which

com-pletely

informationally

some

deep within the bowels

Weights

and

Here is what I
mean by this.

problem

measurement

for

the

kilogram and the standard

carefully guarded vault

of

as a probabilistic

to

quantum

standard

That is, a measurement

standard

of it

respect

with

the

specify

finding

of

unique

density

are

called

measurements

complete

probabilities

and have been studied

time [60, 61, 62]. Here however, the

picture is most pleasing if

we

.
.

consider

refined version of the notionthat

a slightly

of the minimal

mea-surement [32] The

suremen
t [32].
space of Hermitian operators on
HD is
itselfTh
a
2
linear
spacevector
of Herm
space
it ian
of opera
dimension
tors on
D HD
The quantity
B)
tr(A
serves as an inner product on that space.
Hence, if we can find a POVM
E = {Ed }
informationally

consisting

of D

complete

linearly independent

the probabilities P(d)

=tr(Ed

)now

operators,
thought of

as projections

in the directions of the vectors Ed

will completely specify the operator

. Any two

distinct density operators and must give rise

to

distinct

measurement
outcomes

informationally

Do

outcome

statistics

for

minimal

number

The

can

POVM

have
2.

and

this

still

of

be

complete is D

minimal

informationally

complete

exist? The answer is yes. Here is a

way to produce one, though there are
many other ways. Start with a complete
ere
orthonormal
a re manybasis
othe r
|ejw i
ays.
on HD
St art
Itwith
is easy
a com
to check
plete
2
that the following D on
rank-1 projectors d form
a linearly independent set.
POVMs

simple

= 1,...,D, let

1.For d

= |ej

ihej |,

(33)
where j,too, runs

2.For d

12

= D+ 1,...,

over
12

the values 1,...,D.

D(D + 1),let

(|ej i
+ |ek i)(hej |
+ hek |)

(34)
where j < k.

3.Finally, for d

12

12

D(D + 1) + 1,...,D

(|ej i
+ i|ek i)(hej

|
ihek

2,
let

|)

(35)
where again j < k.

All that remains is to transform these

(positive-semidefinite) linearly independent

operators d into

a proper

POVM. This

can be

done by considering the positive semidefinite

operator G defined
by

(36)
d=1

22

This idea has its roots inL.Hardys two important

papers

Refs. [58] and [59].

21

to show that h|G|i

It is straightforward

positive

linear

1/2

(36), we find
D

G is

with positive

transformation

to

a valid decomposition

1/2d

Eq.

that

and hence invertible. Applying the

(invertible)
1/2XG

establishing

(i.e., Hermitian

definite

eigenvalues)

0, thus

for all |i

>

1/2

of the identity,

(37)
d=1

The operators
1/2d
Ed
G

1/2

(38)
satisfy the conditions

(7), and

linear

moreover,

independence

of

a POVM,

they

Eqs. (6) and

retain the rank

of the original d

and

Thus

we

have what

we need.

With

the

existence

of

minimal

informationally
complete POVMs assured, we
can think about the vault in Paris. Let us
suppose from here out that it contains a machine
that enacts a minimal informationally complete
POVM Eh whenever it is used. We reserve the
index h to denote the out-comes of this standard

quantum measurement

for they will replace the

notion of the hypothesis

theory.

Let

us

develop

in classical statistical

this from

Bayesian

point of view.
Whenever

one

has

a quantum system

in

mind, it is legitimate for him to use all he knows

and

believes

of it

to ascribe

function P(h) to the (potential)

this standard

probability

outcomes

of

measurement. In fact, that is all a

quantum state is from this point of view: It is

a subjective

judgment

will obtain

as a result

one

s system

performs

systemone

about which

of

an interaction

and that machine.

measurement
different

consequence
between

Whenever

one

{Ed }
andon that
themac

from

the

standard

quantum measurement
level

{Eh }at the most basic

understanding,

of

gathering

all

(or evoking)

(among other things) allows

initial

ones

subjective

something else Pd (h).

a piece

is not of information
If

completely

integer
free

to

already

be

supposes

for which the standard

2

device has D

outcomes

one is no longer
just any subjective
There are constraints.

D), then

to

make

us call the allowed

PSQM

we may

mechanics and

judgment P(h) he pleases.

Let

P(h)

theoretic origin. It is this.

measurement

some

from

of quantum mechanics that

one accepts quantum

one has a system

(for

is

to recognize is that, with

that

quantum

doing

one to update

judgment

of description,

edging toward

is

23

What is important

this change

one

piece of data d that

region of initial judgments

For instance, take the POVM in Eq. (38)

as

the standard quantum measurement. (And thus,

now

label its outcomes

Then,
from

one can show

by h rather than d.)

that P(h) is bounded

unity, regardless

away

of ones initial quantum

state for the system. Inparticular,

P(h)

max

tr(Eh )

max

tr(Eh )

max

(Eh )

max

(G

max

(G

1/2h
1)

1/2)

= max (h

1h

(39)

where the second line above refers to a

maximization over all one dimensional projectors
and max () denotes the largest eigenvalue of its

argument. On the other hand,

the eigenvalues

of G

one can calculate

explicitly.[63] Through

this, one obtains

?1

P(h)

12

+ cot

3
4D

<1

(40)
23

We will come back to describing the precise form of

this update and its similarity

to Bayes rule inSection 6.

22

Figure 1: The planar surface represents the

of all probability distributions

quantum

Accepting
accepting

that ones

outcomes

of

mechanics
subjective

standard

point

within

the

is,
beliefs

in

space

outcomes

part,

for the

quantum measurement

device will not fall outside

Each

over

a certain convex set.

region represents
a

perfectly valid quantum state.

For large D, this bound asymptotes

1.

to roughly

(0.79D)

More

generically,

informationally

be bounded

for

any

minimal

complete POVM {Eh }, P(h) must

away

outcomes.

Thus

something

driving

mechanics

and

from unity for all its possible

even at this stage, there is

a wedge between quantum
simple

Bayesian

probability

one accepts quantum mechanics,

one voluntarily accepts a restriction on ones
subjective judgments for the consequences of a
theory. When

standard
For all

quantum

consequences

measurement
h, there

intervention:

are no

conditions

whatsoever convincing enough to compel one to

a probability ascription P(h) = 1. That is, one
gives up on the hope of certainty. This, indeed,
one might pinpoint as an assumption about the
physical world that goes beyond pure probability
theory.

24

But what
terms?

What

is that
is

our

assumption
best

in physical

description

of the

wedge?
Some think they already know the answer,

25

Wither Entanglement?

and it is quantum entanglement.

When two systems, of which

we know

the states by their

re-

spective representatives, enter into temporary physical inter-

action due to known forces between them, and when after

time of mutual influence the systems separate again, then they

can no longer

be described in the

endowing each of them with

would not call that

one

quantum mechanics, the

same way as before,

a representative

of its

viz. by

own.

but rather the characteristic trait of

one that

enforces its entire departure

rep-

from classical lines of thought. By the interaction the two

resentatives (or -functions)

have become entangled.

Erwin Schr odinger, 1935

24

It

is

at

this

point

quantum mechanics
and

sees

[59]. Hardy

generalization

and

quantum

light of

a certain

how

of

mechanics

to probability

takes into account

the

present

most crucially

extension

theory, whereas quantum

restriction

that

differs

mechanics

classical

of

as

probability

is depicted here

theory. It is

we

account

from Refs. [58]

restriction

as a
that

ought to think and gamble

physical facta fact

we are

in

working like

crazy to identify.
25

This is not

a spelling

mistake.

23
entanglement

Quantum

captured

the attention of

ac-counts

most

quantum

it

is the

information

computing [64]

has

our

certainly

community.

main

theory

ingredient

and

By

in

quantum

and it is the main mystery of the

quantum foundations [65]. But what is it? Where

does it come from?
The predominant
this
For

paper has
it, more

been
than

quantum mechanics,
information

purpose it has served in

as a kind of background.
any other ingredient in
has clinched

the issue of

about what? inthe authors mind:

a preexisting
we are willing

That information cannot be about

reality (a hidden variable) unless

to

renege on our reason

quantum
place.

states

What

objective

for

rejecting

reality

I
am alluding

the

in the first

to here

is the

conjunction of the Einstein argument reported in

Section

3 and

inequality

the

violations

phenomena
by

quantum

of

the

Bell

mechanics.

gave us that the

me
information
symbolized
by a |i must
be
hat
information
the information
about the
symbolized
potential consequences of
our interventions into the world.
But, now I
would like to turn the tables
and ask whether
the structure
of our
potential interventionsthe POVMscan tell us
Putting those points together

something

about

the origin of entanglement.

Could it be that the concept of entanglement

just

that

a minor addition to the much deeper

mea-surements have this structure?

The technical

we

is, why do

the

tensor-product

and

take the direct

HB

HAB

sum

= HA

= HA

certainly

third HAB

of the two

spaces
r

spaces

We could

HAB

= HA

product

ssmascads
nnpro
of

coul
HB dtake
[66]
We
t hei
could
Gratake

But instead

product,
HAB

are

We could take their Grassmann

other things.

(41)

There

combine two Hilbert

to obtain

HB

of this question

systems according to

rule?

ways to

innumerable
HA

translation

combine

is

point

HB

we

take their tensor

Why?

it

Could

arise

as

considerations

sectionnamely,

property

the

the

of

from

previous

noncontextuality

The

answer

is

theorem I
am about demonstrate

inspiration to Ref. [67].

yes, and the

owes much in

26

Here is the scenario. Suppose

quantum

systems,

measurement

on

selfsame

measurement-outcome

for

probabilities?

from

and

we can

we

have two

a
we can

make

each. On the first,

measure any POVM on the DA -dimensional

Hilbert space HA ;on the second, we can
measure any POVM on the DB -dimensional
Hilbert space HB
(This, one might think, is
the very essence of having two systems rather
than
onei.e.,
that
we can probe them
independently.)
Moreover,
suppose we may
condition the second measurement on the nature
and the outcome of the first, and vice versa.

That is to saywalking
first
the

measure
outcome

from A to Bwe could

{Ei } on A, and then, depending

i,

i
meas ur e{F
{ Ei
}on
and
measure
}
onA,B.
j

on
then

Similarlywalking

measure

{Fj } on B, and then, depending

outcome j,measure

Ei
Xi

{E

we must

valid POVMs,
Xi

from B to Awe could first

} on A. So that
i
have

Xj

=I

and
X

Fj

=I

on

we

the

have

(42)

and
Xi

=I

Xj

and

Fj

=I

(43)

for these sets of operators. Let us denote by Sij

an ordered pair of operators, either of the form
j
i
(Ei ,Fj )or of the form (E i,Fj ),as appearing
above. Let us call a set of such operators {Sij }
a locally-measurable POVM tree.
26

After posting Ref. [1],Howard Barnum and Alex Wilce

brought to my attention that there is a significant amount

of literature inthe quantum logic community devoted to

similar

ways

of motivating the tensor-product

rule. See

for example Ref. [68] and the many citations therein.

24

now

Suppose

POVM-version

thatjust

of Gleasons

as

with

the

theorem in Section 4

the joint probability P(i,j)for the outcomes of

such

a measurement

should

upon

not depend

which tree Sij is embedded in:This is essentially

same

the

assumption

applied to local

we

systems. In other words, let

exists

a function

f :EDA

EDB

(44)

now
on the separate
us suppose there

made there, but

measurements

[0,1]

such that
Xij

f(Sij )

Xij

(45)

whenever

the Sij

or

Eq.

that Eq. (44) makes

no

satisfy either Eq. (42)

(43).
Note in particular

use

of the tensor product: The domain of f is

the Cartesian product of the two sets EDA

and EDB

The

notion

of

local

on

measurement

the

separate systems is enforced by the requirement

that

the ordered

conditions

course,

pairs

of Eqs. (42)

satisfy

Sij

and

(43).

the

side

This, of

is not the most general kind of local

one

measurement
sophisticated

can

imaginemore

measurements

multiple ping-pongings

the present

Ref. [69]but

could

involve

as

in

class

is

between A and B
restricted

already sufficient for fixing that the probability

rule for local measurements

must

come

from

tensor-product structure.
The theorem

27

is this: If f satisfies Eqs. (44)

and (45) for all locally-measurable

then there exists
HB

linear operator

defined

over

such that
L(E

f(E,F)

= tr

F)

and HB

are

POVM trees,
L

on

HA

(46)
If HA

the field of

complex numbers, then L is unique. Uniqueness

does
complex
not numbers
hold, however, if the underlying field is
the real numbers.

The proof

of this

statement

is almost

trivial extension of the proof in Section 4. One

, ,

again starts by showing additivity, but this time

in the two variables

instance, for
gE

(F)

a fixed

E and F separately.

E EDA

f(E, F)

(47)

For

define

and consider two locally-measurable POVM trees

{(I E,Fi ),(E,G )}

(E,H )}

and

{(I E,Fi ),

(48)

where {Fi },{G },and {H } are arbitrary

POVMs

on HB

Xi

Then Eq. (45) requires that

gI -E

(Fi )+

=1

gE

(G )

gE

(H )

(49)

and
Xi

gI -E

(Fi )+
X

(50)

From this it follows that,

=1

gE

(G )

gE

=const.

(H )

(51)

That is to say, gE (F) is a frame function inthe

sense of Section 4.Consequently, we know that

we can use the same methods as there to uniquely
extend gE (F) to a linear functional on the
complete set of Hermitian operators on HB
Similarly, for fixed F EDB we can define
hF

(E)

=f(E,F)

(52)
27In

,
,

Ref. [1], a significantly

stronger claim is made:

Namely, that L
is in
fact
a density operator.
This
Ref.
[1],
a significantly
stronger
flat-out

was a

mistake. See further discussion below.

25
and

prove

a linear
on HA

uniquely to

operators

functional

The linear extensions

can

can be extended
on the Hermitian

that this function too

of gE (F) and hF (E)

be put together in a simple

way to give a

full bilinear extension to the function f(E, F).

Namely, for

any two

Hermitian operators A and

B on

HA

2 E2

and B

such that

EDB

F2

and HB

= 1

respectively, let A

F1 2 F2

1 ,2 ,1 ,2

gE1

(B)

gE2

(53)

see

To

EDA

(B)

= 1

E1

2 E2

(54)
Then

we have
f(A,B)

=
=

=
=

1 hF1 (A)
1
1
1
1

1 g E1
(B) 2 g E2

(B)

,B) 2 f(E2
,B)
1 f( E1

,F1)
1 1 f( E1,F1
) 2 f( E2

,F2 )
2 1 f( E1,F2
) 2 f( E2

hF2

(A)

f(E1,F1 ) 2 f(E2

f(E1,F2 ) 2 f(E2

,F1)

,F2

f(E1 ,B) 2 f(E2 ,B)

and F1

that this definition is unique, take

other decomposition
A

,.

E1

be decompositions

0, E1,E2

Then define

f(A, B)

= 1

any

1 gE1 (B)

gE2

(B)

(55)

which is as desired.

,
, .
,.

With bilinearity for the function f

we have essentially

established,

70]. For, let {Ei },i 1,...,D

the full story[66,

be

a complete

on HA and
a complete basis for

basis for the Hermitian operators

=1,...,D

let {Fj }, j

be

the Hermitian operators

Pi

i Ei

f(E,F)

and F

Xij

Pj

j Fj

i j

on HB

If E

then

f(Ei ,Fj )

(56)
Define L to be a linear operator

satisfying the (DA

L(Ei

tr

Fj

DB

= f(Ei

on HA

HB

linear equations

,Fj

(57)
Such

have,

an operator

always exists. Consequently

we

f(E,F)

i j

L(E

tr

F)

L(Ei

Xij

tr

Fj

(58)

For complex Hilbert

spaces

HA and HB

uniqueness of L follows because the set {Ei

the

Fj }

a complete basis for the Hermitian

operators on HA HB .[71] For real Hilbert
spaces, however, the analog of the Hermitian
operators are the symmetric operators. The
dimensionality of the space of symmetric
12
operators on a real Hilbert space HD is D(D +
2
1),rather than the D it is for the complex case.
This means that inthe steps above only
forms

14
DA DB

(DA

+ 1)(DB + 1)

(59)

26
equations will appear inEq. (57), whereas

12
DA DB

(DA

+ 1)

DB

(60)

are

an

L. For

= DB = 2. Then

Eq. (59)

to uniquely

needed

instance

take DA

specify

gives nine equations, while Eq. (60) requires ten.

This establishes

nice if
full

we

could

probability

the theorem. It would be

go

further and establish the

rule

measurementsi.e.,

our

operator. Unfor-tunately,
strong

enough

counterexample

for

LS

(E F)

assumptions

density

are not

is a
a linear operator
to the swap operator on the
that.

spaces:

quantum
be

Here

[72] Consider

that is proportional

two Hilbert

local

for

that L must

F E
D

(61)

This

clearly

satisfies

the

conditions

of

our

theorem, but it is not equivalent

to

density

operator.
Of

course, one

by requiring

that

could

recover

it give

even for nonlocal measurements

ven
of the identity operator on HA
the

positivity

positive

(i.e., resolutions

HB ).But in

purely local setting contemplated

would be a cheap

good

way out. For, one

here, that

should ask in

what ought to be the rule for

conscience

defining the full class of measurements

nonlocal

for L

probabilities

:Why

measurements)

(including
should

it

to an arbitrary resolution of the

on the tensor product? There is nothing
makes it obviously so, unless one has

correspond
identity
that

already accepted standard

Alternatively,

quantum mechanics.

it must be possible to give

purely local condition that will restrict L to be

density

operator.

noted
L to beabove,

function

This

is uniquely

f(E, F); we

never

the
functi
than
theon
probabilities

outcomes in specifying
condition supplies

is because
determined

L,
by

as

the

need to look further

of local measurements

L. Ferreting

an avenue

out such

for future research.

All of this

L is, it must

away
be a

the tensor product of HA

and

does

not, however,

from the fact that whatever

HB

on

operator

linear

Therefore, let

striking

feature

tensor-product

us

close by emphasizing

way

of this
rule

take

deriving

of

the

separate

combining

for

quantum systems: It is built

the

on the very concept

of local measurement. There is nothing spooky

or

nonlocal

about

it; there is nothing in it

resembling passion at

one

did not

even

have

a distance

for the outcomes

assignments

[73]

Indeed,

to consider probability
of measurements

of the nonlocality without entanglement

[69] to uniquely fix the probability

isto give

an example on

H3

H3

not consider standard measurements

|d

|1
|2

ihd

|},d

variety

rule. That

one need
like {Ed

= 1,...,9, where

=
i
= |0i|0 + 1i
i |1i|1i

|6

i |1+

|7

i |1

2i|0i
|3

i |0i|0 1i

2i|0i

(62)

|4

i |2i|1+ 2i

=
=

+ 1i|2i

|8

i |0

|5

i |2i|1 2i

1i|2i

|9

an

with |0i, |1i, and |2i forming

basis

on

H3

and |0

1i

1
2

i |0

orthonormal

(|0i

|1i), etc.

a measurement that takes neither the

form of Eq. (42) nor (43). It stands out instead,
inthat even though all its POVM elements are
tensor-product
operatorsi.e.,
they have no
This

is

quantum

entanglementit

measured

by

elaborate

ping-ponging

local

still

means,

cannot

be

with

the

even

strategies

mentioned

earlier.

Thus, the tensor-product

rule, and with it

seems to

more a

quantum

entanglement,

statement

of locality than anything else. It, like

the probability

rule, is

more a

be

product

of the

structure of the
27
observablesthat
with

they

noncontextuality.

are

POVMscombined

In searching

for

the

secret

to

ingredient

general

Bayesian

quantum mechanics,

drive

wedge

probability

it

not to look is toward

seems

between

theory

and

that the direction

quantum

entanglement.

Perhaps the trick instead is to dig deeper into the

Bayesian toolbox.

Whither Bayes Rule?

And

so you see

as true

I
have

come to doubt

I
stand alone without beliefs

28

Allthat I
once held
The only truth I
know

is you.
Paul Simon,

timeless

states

Quantum

knowledge,

are states

not

commitments,

of information,

pragmatic

belief,

states

of

gambling

nature.

That

paper. Thus,
sense of the remainder of
one strategy ought to be to

statement is the cornerstone of this

in searching to make

quantum mechanics,
seek

guidance

avenue

of

Bayesian

[74] from

probability

most

the

rational-decision

theory

theory

[75,

developed

to date
76,

77]

Indeed, the

very

develop

reliable

making

decisions

aim of Bayesian theory is to

methods

reasoning

of

in the light

and

of incomplete

information. To what extent does that structure

mesh with the seemingly
of

quantum

mechanics?

independent
To what

structure

extent

are

there analogies; to what extent distinctions?

a distinction into
an analogy. The core of the matter is the manner
in which states of belief are updated in the two
theories. At first sight, they appear to be quite
different in character. To see this, let us first
This section is about turning

explore

how quantum mechanical

states change

when information is gathered.

In older

accounts

of quantum

mechanics,

one often encounters the collapse postulate as

a basic statement of the theory. One hears things
like, Axiom 5: Upon the completion of an ideal
measurement of an Hermitian operator H, the
system is left in an eigenstate of H. In
quantum

information,

however,

it has become

clear that it is useful to broaden the notion of

measurement,

and with it,the analysis of how

can change in the process. The foremost

reason for this is that the collapse postulate is
simply not true ingeneral: Depending upon the
state

exact

nature

may

the measurement

of

interaction,

any of a large set of possibilities

the final state of a system.
there

The
change

be

broadest
arises

opera-tions

consistent

in the theory

[50]

operator

measured

on

of state

of effects

and

The statement is this. Suppose

that

a quantum system is a
a POVM {Ed } is
system.
opera torThen,
an d
according
a POVM {E
to

ones initial state for

density

notion

for

and

this formalism, the

state after the measurement

can be any state d

of the form

Xi

di

Adi A

tr(Ed )

(63)
where
Xi

di

(64)

Adi

=Ed

Note the immense generality of this formula.

There is no constraint

on the number

of indices i

in the Adi and these operators need not

even be

Hermitian.
28

This is not

a spelling

mistake.

28
The usual

justification

as

generalityjust
commonplace

than

as a

arises in

an

an

of

the

ancilla

of

the

POVM

that the

indirect fashion rather

direct and immediate

with

case

for

about by imagining

other words, the primary

interact

the

justification

formalismcomes

measurement

in

for this kind

observation.

In

system is pictured to
first, and

only

then

a real measurement on the ancilla

one posits a kind of
projection postulate on the primary system due
to this process. This assumption has a much safer
feel than the raw projection postulate since, after
the interaction, no measurement on the ancilla
should cause a physical perturbation
to the
subjected to

alone. The trick is that

primary system.

More

formally,

we can start

out

by

following Eqs. (9) and (10), but inplace of Eq.

(11)

we must

make

an assumption on how the

For this one invokes a

systems state changes.

kind

of

Namely,

projection-postulate-at-a-distance.

one takes

(I

d )U(S

trA

)U

(I

d )

P(d)

(65)

reason for invoking the partial trace is to

make sure that any hint of a state change for

The

the ancilla remains unaddressed.

To see how expression (65) makes connection

to Eq. (63), denote the eigenvalues and

of A by and |a i
respectively.

eigenvectors
Then S
S A

(66)

can be written as

|a iS ha |

and, expanding Eq. (65), we have

29

ha

(S

|(I d )U

)U(I d )|a

P(d)

p
ha

|(I d )U

P(d)

ha

|U(I d )|a

|a

(67)
A representation of the form inEq. (63)

can be

made by taking

Ab

ha

|U(I d )|a

(68)

and lumping the two indices and into the

single index
that

i.Indeed,

Eq. (64) holds.

30

one can

easily

had set out to show. However, just

check

what

we

as with

the

This completes

case of the POVM {Ed }, one can always find a

way to reverse engineer the derivation: Given a

set of Adi

one can always

find

a U, a A

of d such that Eq. (65) becomes true.

course

Of

contained

There,

{Ed }and {Adi

projection

old

collapse

and set

postulate

within the extended formalism

case:

special

the

one

takes

both sets

= Ed } to be sets of orthogonal
us
case

operators. Let

al
think
projection
about this special
is distinctive

just

is

as a

take

a moment to

in isolation. What

about it is that it captures in the

a common folklore associated with the

measurement process. For it tends to convey the
image that measurement
is a kind of
gut-wrenching violence: In one moment the state
is = |ih|, while in the very
In one
nextmoment
it is a i
the
=
|iihi|.
Moreover, such a wild transition need
depend upon no details of |i and |ii; in
particular
the
two states
may even
wild transition need depend upon no d
extreme

29

David Mermin has also recently emphasized

this point

inRef. [78].
30

As

an aside,

it should be clear from the construction in

Eq. (68) that there

of d

For

are many

precise

equally good representations

statement

of the

latitude

of

this

freedom,

see

Ref. [79].

29
be almost orthogonal to each other. In
density-operator language, there is no sense in
which i is contained in : the two states

distinct places of the operator

Xi

6=

P(i)i

space.

are in

That is,

(69)
Contrast this with the description of

information gathering that arises inBayesian

probability

theory. There,

belief is captured by

an initial state of

a probability

P(h) for some hypothesis H.The

a piece

distribution

way gathering

of data d is taken into account in

assigning ones

new state of belief

is through

Bayes conditionalization rule. That is to say,

one expands

P(h) in terms of the relevant joint

probability distribution and picks off the

appropriate term:

P(h)

Xd

P(h,d)

Xd

P(d)P(h|d)

(70)

P(h)

P(h|d)

(71)

where P(h|d) satisfies the tautology

P(h|d)

(72)
How

P(h,d)

P(d)

gentle

quantum

this

collapse!

information,

one

looks

in

When

comparison

one

gathers

to

new

simply refines ones old beliefs

inthe most literal of

senses.

new state is incommensurable

was always there; it was just

It is not

as if the

with the old. It

initially averaged

inwith various other potential beliefs.

Why does quantum

collapse

not look

more

rule? Is quantum

like Bayes

more

collapse

really

a
an

or might it be
a problematic representation? By this
stage, it should come as no surprise to the reader
that dropping the ancilla from our image of
violent kind of change,

artifact of

measurements will be the first step to

generalized

progress.

Taking the transition from to d

as the basic statement

measurement
is is a good

in

Eqs. (63) and (64)

of what

quantum

starting

point.

To accentuate

similarity

between Eq. (63)

us first contemplate cases of

i
takes on a single value. Then,

and Bayes rule, let

it where the index

we can conveniently
1

(73)
where

Ed

Ad

(74)
In a loose

given

way, one can say that measurements

are the most efficient they can be for
POVM {Ed }: For, a measurement

of this sort

Ad A

P(d)

drop that index and write

interaction with

viewed

as

more

away some

i-dependence

a measurement

truly

of the information

make this point

more

be

of

that ones uncertainty

us

rule has the property

about

be expected to decrease

can be

it gained. Let

precise.

that Bayes

data. This

may

POVM that just happens to throw

finer-grained

Notice

an explicit

upon

can

hypothesis

the acquisition of

made rigorous, for instance,

by gauging uncertainty with the Shannon entropy

function [80],
S(H)

Xh

P(h)logP(h)

(75)

30
This

number

is bounded

between

0 and

the

logarithm of the number of hypotheses in H, and

are several reasons to think of

measure of uncertainty. Perhaps
there

important

of

these

is that

number of binary-valued

it

as a good
the

it quantifies

questions

most

the

one expects

to ask (per instance of H) if ones only

ascertain the outcome is from
knows the result [81]

the

lower

predictable

the

means to

colleague who

Under this quantification,

entropy,

Shannon

more

the

a measurements outcomes.

Because

the

function

concave on the interval

f(x)

xlogx

is

[0,1], it follows that,

S(H)

Xd

Xd

Xh

P(d)P(h|d)

Xd

log

P(d)P(h|d)

Xh
X

P(d)

P(h|d)logP(h|d)

Xd

P(d)S(H|d)
X

(76)

Indeed

we

hope to find

a similar statement
measureor impredictability.
by a decrease of

for how the result of efficient quantum

ments
But,

decrease
what

can

uncertainty
have

argued

uncertainty
be

through

meant

quantum

strenuously

that

measurement?

the information

gain in
about

a measure- ment cannot

be information

preexisting

way out

reality. The

impasse is simple: The uncertainty

in quantum measurement

expects

for

the

of the

that decreases

one

is the uncertainty

results

of

other

potential

measurements.

are at least two ways of quantifying this

are worthy of note. The first has to do with
von Neumann entropy of a density operator

There
that
the

:
D

S() X trlog

(77)

logk

k=1

where the k

signify the eigenvalues

of . (We

use the convention that log = 0 whenever =

0 so that S( ) is always well defined.)
The intuitive meaning of the von Neumann
entropy can be found by first thinking about
the
Shannon
entropy.
Consider
any von
Neumann

measurement

one-dimensional

Shannon

P consisting

orthogonal

entropy

for

the

projectors

outcomes

.
of

of

The

this

measurement is given by

=
H(P) =
(78)

Xi=
Xi=1

H(P)

(tri )log(tri )
Xi=1

A natural question to ask is: With respect

given density

operator

will give the most predictability

will
As it turns out, the

set

answer

of eigenprojectors

orms
obtains,
a setthe
outcomes

entropy

is

over its outcomes?

any P that forms a

for [82]

When

this

Shannon entropy of the measurement

reduces to simply the

of

to

, which measurement

the

density

Neumann entropy,

von

operator.

Neumann

The

von

then, signifies the amount of

way of a
standard measurement in a best case scenario.
Indeed, true to ones intuition, one has the most
predictability by this account when is a pure
statefor then S() = 0. Alternatively, one has
impredictability

one

achieves

by

the least knowledge when is proportional to the

identity

operatorfor

will have outcomes that

any measurement
are all equally likely.

then

The

best

however,

is

indicative

of

case scenario for predictability,

a limited case, and not very
the density operator as a whole.

Since the density operator contains, inprinciple,

all that

can

said about every possible

seems a shame to throw away
part of that information
in our
be

measurement, it
the

vast

considerations.

31
This leads to a second method for quantifying

we
as our

uncertainty

in the quantum setting. For this,

again rely

on

the Shannon

information

basic notion of impredictability. The difference is

we

evaluate

measurement

it

with

respect

to

typical

rather than the best possible

one.

But typical with respect to what? The notion of

typical is only defined with respect to a given

measure on the set of measurements.
Regardless, there is a fairly canonical answer.
There is a unique measure d on the space of
one-dimensional

projectors

that

is

invariant

with respect to all unitary operations. That in

turn induces

canonical

measure

dP

on

the

space

of

Using this

von Neumann measurements

measure leads to the following

P [83]

quantity

S()

H()dP

(tr)log(tr)d

(79)
which is intimately connected to the so-called

quantum subentropy of Ref. [84].

entropy

can be evaluated
?12

on the expression

13

+ +

ln2

+ Q()

(80)
where the subentropy Q() is defined by
k

Q()

k=1

i6=k

logk
k

mean

explicitly in terms of

the eigenvalues of and takes

S()

This

(81)
In the

case

where has degenerate

= m

+ and m

em
0. to
The

for l6=

m, one

eigenvalues,

need only reset them to

and consider the limit

limit is convergent

and hence

finite for all . With this, one

Q()

is

as

can also see that

finite
for a pure state , Q() vanishes. Furthermore,

since

S() is bounded above by logD,

we

know

that
?12

1
0 Q() logD

+ +

ln2

ln2

(82)

where is Eulers constant. This

any

, Q()

never

means

exceeds approximately

that for

0.60995

bits.
The

interpretation

of

this

result

is the

when
one has maximal
a quantum systemi.e., one
has a pure state for itone can predict almost
nothing
about
the outcome
of a typical
following.

information

measurement

Even

about

[40]

In the limit of large d, the

a typical measurement is just

over a half bit away from its maximal
Having a mixed state for a system,
ones predictability even further, but

outcome entropy for

a little
value.
reduces

indeed not by that much: The small deviation is

captured

by the function

a quantification

becomes

in Eq. (81), which

of uncertainty in its

own

right.

way to get at a quantum statement of

use of the fact that S()
and Q() are both concave in the variable . [85]
That is,for either function, we have
The

Eq. (76) is to make

F(t 0

+ (1

t) 1 ) tF( 0 ) + (1 t)F( 1 )

(83)
for any density operators 0 and 1 and

number t [0,1]. Therefore,

Xd

F()

F()
(84)

32

P(d)F(d )

one

any

real

might hope that

Such a result howeverif it is truecannot

inthe trivial fashion it did for the classical

arise

case

of Eq. (76). This is because generally (as already

emphasized),
Xd

6=

P(d)d

(85)

as inEq. (73). One therefore must

for d defined

be more roundabout if a proof is going to happen.

The key is innoticing that

1/2

1/2

Xd
1/2

Ed

1/2

Xd

P(d) d

(86)

where

1/2

P(d)

Ed

1/2

1/2

Ad

P(d)

1/2

(87)

What is special about this decomposition

of is

that

for each d, d

eigenvalues.

have the

same

In the present

have the same

X
and XX

and d

This follows

since X

eigenvalues, for

case, setting

any operator

= Ad

1/2

X.

does the

trick. Using the fact that both S() and Q()

depend only

upon

the eigenvalues of we obtain:

Xd

S()

P(d)S(d )

(88)
Xd

Q()

P(d)Q(d )

(89)

as we

had

an

performing
of

been

a POVM

{Ed

hoping

efficient

for. Thus,

quantum

}, an observer

in

measurement

can expect to

left with less uncertainty than he started with.

In this
have

sense, quantum

some of the flavor

more, and the

expect

does indeed

of Bayes rule. But

we can

derivation above hints at

just the right ingredient:

same

collapse

be

31

and d

have the

eigenvalues! To see the impact of this, let

us

once

again explore the content of Eqs. (73) and

(74). A

common way to describe

their meaning is

to use the operator polar-decomposition theorem

[87] to rewrite Eq. (73) inthe form

1/2

Ud E

P(d)

1/2

(90)

where Ud
only

to

is

a unitary operator.

the

constraint

Sincesubject

of

efficiencythe

are not determined any further

than Eq. (74), Ud can be any unitary operator
whatsoever. Thus, a customary way of thinking of
the state-change process is to break it up into two
conceptual pieces. First there is a raw collapse:
operators

Ad

1/2

P(d)

1/2

(91)

Then,

subject

measurement

to

the

interaction

details

and

of

the

the

particular

outcome d, one imagines the measuring device

enforcing
feedback

a further kind
on the measured

of back-action

system[88]:

or

Ud

d U

(92)
31

By differing

terms of

methods,

a strengthening

of this result in

a majorization property can be found

in Refs. [85]

and [86].

33
But this breakdown of the transition is a purely

game.
the Ud are arbitrary to begin with, we

conceptual

Since

as well break down the state-change

process into the following (nonstandard)
conceptual components. First one imagines an
might

observer refining his initial state of belief and

simply plucking out

a term corresponding to the

data collected:

Xd

P(d) d

(93)

(94)

Finally,

may

there

takes

account

into

be

measurement

details

interaction

beliefs, which

both

and

initial quantum state. This

mental

further

of the observers

readjustment

the

of

the

observers

some

is enacted via

(formal) unitary operation Vd :

Vd d

(95)
Putting the two

same

result

processes

together,

one

has the

as the usual picture.

The resemblance

between the

process

in Eq.

(94) and the classical Bayes rule of Eq. (71) is

uncanny.

32

By

this

way

of

viewing

quantum collapse is indeed not such

things,

violent

state of affairs after all. Quantum measurement is

more,

nothing

refinement

and

and

nothing

readjustment

less,

than

of ones

initial

state of belief. More general state changes of the

form Eq. (63)
further

away

come

about similarly, but with

step of coarse-graining

information that

was

(i.e.

throwing

inprinciple accessible)

us

Let

maximally

account,
it.This

1/2

Ed

In the

first,

whose initial belief structure

observer

is

cases of efficient
we imagine an

look at two limiting

measurements.

tight

|ih|

state of belief. By this

no measurement whatsoever can refine

follows because, no matter what {Ed }is,

1/2

P(d)|ih|

(96)
The only state

from

change

a measurement

we just change
consequence
of
is

a sense

can come

must be purely

about

of the

new;
as a
the side effects
of our
intervention. That is to say, there

mental-readjustment

experimental

that

sort: We learn nothing

what

we can

in which the

predict

measurement

is solely

disturbance. In particular, when the POVM is

orthogonal set of projectors

state-change

mechanism

{i

collapse postulate, this simply

readjustment

(97)

von

and the

Neumann

corresponds

to

according to unitary operators Ui

whose action inthe subspace

|
iih |

= |iihi|}

is the

an

spanned by |i is

At

the

opposite

at

all

that

causing

of

we can
have
no

of things,

measurements

contemplate
possibility

end

physical

disturbance to the system being measured. This

come about, for instance, by interacting

with one side of an entangled pair of systems and
using the consequence of that intervention to

could

ones beliefs about the other side. In

update
such

a case, one can show

is purely

mental

further

AB

decomposition takes the form

|

AB

Xi

variety
33

readjustment).

a pure state

consider

that the state change

of the refinement

pi

|ai i|bi i

(with

For

no

instance,

iwhose

Schmidt

(98)

32

Other

Bayesian

similarities

between

conditionalization

quantum

collapse

have been discussed

and

in Refs.

[89, 90, 91].

33

This should be contrasted

minimally

disturbing

our case, a

minimal

with the usual picture of

measurement
disturbance

of

some

version

of

POVM. In

POVM

{Ed }corresponds
In

the

usual

corresponds

to taking Vd

presentationsee

to taking

Ud

=I
for all d in Eq. (95)
Refs.

[85]

and

[88]it

I
for all d in Eq. (92)

instead. For

34
An efficient measurement on the A side of this
leads to a state update of the form
AB

AB

ih

Tracing out the A side, then gives

Ad

trA

(Ad I)|

Xijk

qj

AB

ih

I|

AB

|(A

AB

I)

AB

ih

|A

=
=
=
=
=

pk

(99)

hai |Ad I|aj i|bj ihak |hbk |A

I|ai i

qj
qj

Xijk
Xijk

pk
pk

hak |A

qj

Xjk

pk

|ai ihai |Ad |aj i|bj ihbk |

hak |A
qj

Xjk

Ad |aj i|bj ihbk |

pk
pk

hbk |UA
qj

Xjk
Xjk

Ad U

UA

Ad U

i|bj ihbk |

T
d

1/2

i|bj ihbk |

T|bk

UA

pk

hbj |

|bj

Ad U

1/2

(100)

where is the initial quantum

side, U is the unitary

state

on

the B

operator connecting

|ai i basis to the |bi i

basis, and

the

represents taking

|ai
i
a transpose
with respect to the |bi ibasis. Since
the operators

Fd
UA

(101)

go

UA

T
d

Ad U

together to form

POVM,

we

indeed have

the claimed result.

In summary, the lesson here is that it turns

easy to think of quantum

as a noncommutative variant of Bayes

out to be rather
collapse

rule. In fact it is just inthis that

get

feel

for

noncontextuality
classical Bayesian

further

partitioned

we

have just

of the transition P(h)

solely by the local probability

P(d). The transition does not

have

for Gleasons

In the setting of

conditionalization

that: The probability

P(h|d) is governed

reason

assumption.

one starts to

rest

the

care

about how

of

the

we

potential

care whether d is
a two outcome set {d, d} or
embedded in a three outcome set,

transitions. That is,it does not

embedded

in

whether it is

{d, e,(d e)}, etc. Similarly with the quantum

case. The probability for a transition

0 cares not whether our refinement

from

to

is of the

form
17

or

P(0)0

of the form

(102)

d=1

as long as
17

P(18)18

P(d)d

= P(0)0

P(d)d

P(18)18

(103)
d=1

instance, Howard Wiseman writes inRef. [88]:

E
The action of

1/2
d

produces the minimum

change inthe system,E

required
by Heisenbergs

be consistent

relation, to

with a measurement giving the

information about the state specified by the probabilities

[Eq. (8)].

The action of [Ud ]represents

additional back-action,
the

system.
reasonably

...

anunnecessary

perturbation of

A back-action evading measurement

defined by the requirement

[d],[Ud ]equals unity (up to a phase factor that

be ignored without loss of generality).

is

that, for all

This of course

can

means

that, from the present point of view, there is no such thing

as a state-independent
measurement. Given
the minimally

one that

notion of min-imally disturbing

an initial state

and
a POVM
{Ed },
imally
disturbing

disturbing measurement

produces

pure Bayesian

interaction is the

updating with

no further

(purely quantum) readjustment.

35
What could be

a simpler

generalization of Bayes

rule?
Indeed, leaning

on

that,

we can restate

discussion of the measurement

problem

the

at the

be-ginning of Section 4 in slightly

more

technical

terms. Go back to the classical setting of Eqs. (70)

and

(72)

where

an agent

has

probability

distribution P(h, d) over two sets of hypotheses.

Marginal- izing

over

the possibilities

for d,

one

obtains the agents initial belief P(h) about the

hypothesis

h. If he gathers

data d, he should

use

an

explicit piece of

Bayes rule to update his

probability about h to P(h|d).

The question is this: Is the transition

P(h|d)

(104)

a mystery we should contend with? If someone

asked for a physical description of this
transition, would we be able to give an
explanation? After all, one value for his true and
always remains true: there is no transition init.
One value for d is true and always remains true:
there is no transition

init.The only

discontinuous transition is inthe belief P(h), and

that presumably is a property of the believers

brain. To put the issue into terms that start to

sound like the quantum measurement

problem, let

us ask: Should we not

have

how the brain works before

34

validity of Bayes rule?

The answer is, Of

a detailed theory of
we can trust inthe

course

not! Bayes

ruleand beyond it all of probability theoryis

a tool

that stands above the details of physics.

George Boole called probability theory

a law of
way

thought [94]. Its calculus specifies the optimal

an agent

should

reason

and make decisions when

faced with incomplete information. Inthis way,

probability theory is a generalization of

35a
Aristotelian logic
tool of thought few would

accept

as being

physical world.

anchored to the details of the

36

As far as Bayesian probability

a classical measurement
any I-know-not-what that induces an

theory is concerned,
simply

is

application of Bayes rule. It is not the task of

probability theory (nor is it solvable within
probability theory) to explain how the

transition Bayes rule signifies

comes

about

within the mind of the agent.

34

This point

was recently

by Rocco Duvenhage inhis

stated much

paper

Inclassical mechanics
nothing strange. It is merely

more

eloquently

Ref. [92]:

a measurement

anevent

obtains information about

is

where the observer

some

physical system.

A measurement therefore changes the observers

information regarding the system. One

What does the change in the observers

mean? What

causes

it? And

can then ask:

information

so on. These

questions correspond to the questions above, but

now

they

seem tautological

rather than mysterious, since

intuitive idea of information tells

inthe observers

new

has received

us that

information simply

is caused by the reception of the

the state of

means

case

say an observer

a classical system,

new

precise measurements

are not

has information regarding

but not necessarily

case, since

possible inpractice).

Now the observer performs

new

...

information.

is no different

complete information (this isthe typical

system to obtain

that he

information, and that the change

We will see that the quantum

Lets

our

the change

a measurement on the

information ...The observers

information after this measurement

then differs

from his information before the measurement. Inother

...

words,
information.

a measurement

disturbs

the observers

The Heisenberg cut. This refers to an imaginary

dividing line between the observer and the system
being observed ...It

information

crosses

can be seen as the place

where

from the system to the observer,

but it leads to the question of where exactly it

should be;where does the observer begin? Inpractice

a problem:

its not really

the cut is.It is merely

It doesnt matter where

a philosophical

question

which isalready present inclassical mechanics,

since inthe classical

case

information also

the system to the observer and

passes

one could

from

again

ask where the observer begins. The Heisenberg cut is

therefore

no more

problematic

inquantum

mechanics than inclassical mechanics.

35

Inaddition to Ref. [76], many further materials

concerning this point of view

can be downloaded

from

the Probability Theory As Extended Logic web site

maintained by G.L.Bretthorst, http://bayes.wustl.

edu/.
36

We have, after all,used simple Aristotelian logic in

making deductions from all our physical theories to date:

from Aristotles physics to quantum mechanics to general

relativity and

even string

theory.

36
The formal similarities
and quantum collapse

finally

cut

the

measurement
telling

us that

may

us how to

knot

of

the

Namely,

it

may

Gordian

problem.

between Bayes rule

be telling

it is simply not

be

a problem at all

Indeed, drawing

two theories,

one

on

the analogies

between the

a spark

is left with

of insight:

perhaps the better part of quantum mechanics

is simply

law of

thought

[56]. Perhaps

the

way
some

structure of the theory denotes the optimal

to

reason

and make

fundamental

decisions in light of

situationa

fundamental

situation

out in a more satisfactory

waiting to be ferreted

fashion.

This much

we

know:

That

it

ismust

situationwhatever
ingredient

have.

Bayesian probability

As already emphasized,

something

to drive

theories.

Probability

general

a structure.

be

an

theory does not

there must be

wedge between the two

theory

alone

is

too

Narrowing the structure will

require input from the world around

6.1

fundamental

us.

Accepting Quantum Mechanics

Looking at the issue from this perspective,

mean to accept
quantum mechanics? Does it mean accepting (in
essence) the existence of an expert whose
let

us

ask: What

does

it

we should

probabilities

we

strive to possess whenever

strive to be maximally rational? [93] The key

to answering

this question

the previous discussion

considerations

comes

the

of

from combining

of Bayes rule with the

standard

quantum-measurement device of Section 4.2. For,

contemplating

this

than

further

will allow

calling

us to go even
collapse
a

quantum

noncommutative variant of Bayes rule.

Consider the description of quantum collapse

in Eqs. (93) through (95) in terms of ones sub-

jective judgments for the outcomes of

quantum measurement
there,

one starts

standard

{Eh }. Using the notation

with an initial judgment

tr(Eh )

(105)
and,

after

Pd

(h)

= tr(d

Eh )

(106)
where
d

Fh

(107)

Vd

Eh Vd

of some
other
up with a final judgment

measurement

observable {Ed },ends

= tr( d

Eh Vd

= tr( d

Fh )

Note that, in general, {Eh } and {Ed } refer to

two entirely
Note t ha t,
different
in genera
POVMs;
l,{Eh }and
the range
{Ed }ref
of
indices h and d need not

their

Also, since
complete

informationally

even

minimal
d

{F h }

POVM,

will

be the

same.

informationally
itself

be

complete for each value of d.

Thus, modulo

a final
the

redefinition

of

measurement

based

precisely

{Eh } is

Bayes

unitary readjustment

standard

on the data

rule in this

gathered,

one has

transition.

This

follows since

Xd

P(d) d

implies
P(h)

Xd

P(d)P(h|d)

where

P(h|d)

= tr( d

Another

Eh )

way of looking at this transition

from the active

or

quantum

is

(108)
(109)
(110)
point of view, i.e., that the axes
of the probability simplex

are held

fixed, while

the state is transformed from P(h|d) to Pd (h).

That

is,writing

Fh

(111)
0

h =1

37

dhh
0

Eh 0

Figure 2:A quantum measurement

is any

I-know-not-what

that generates

an application

of

Bayes rule to ones beliefs for the outcomes

of

standard quantum measurementthat

decomposition

combination
choice

is

a convex
then a final

of the initial state into

of other

states

and

(decided by the world, not the observer)

within that set. Taking into account the idea that

quantum
disturbing

measurements

are

invasive

alters the classical Bayesian

a further outcome-dependent
One can either
think of it

only in introducing
readjustment:

or

picture

as a

passively

quantum
depicted

readjustment

measurement

as a

here)

the

of

or

device,

actively

adjustment

further

standard
(as

to the

posterior state.

where

{Eh

are some
to a

dhh
0

} refers

Pd

(h)

dhh
0

of

the

P(h

original

we

measurement,

and

coefficients

relabeling

quantum

standard

real-valued

get

0|d)

(112)
0

h =1

This gives

an

enticingly simple description of

what quantum measurement

Modulo

the

measurement
whatsoever

is

on

is in Bayesian terms.

a quantum

readjustment,

final

any

application

of Bayes rule

the initial state P(h).

By

any

application of Bayes rule, I

mean in particular

any convex

decomposition

refinements

P(h|d)

Aside from the final

measurement
measurement:

is

that

also

live

readjustment,

just

It is

of P(h)

like

any

into

some

in PSQM

37

a quantum

classical

I-know-not- what

that

an agent to an

pushes
rule.

application

of Bayes

38

the formal structure

Accepting
mechanics

large

isin

of quantum

.
,

accepting

partsimply

that it would not be in ones best interest to hold

P(h)

that

up to
by a

Moreover,

signified

measurement
37

Note

calling
d),

between this

distribution

should

not

existence

get

that

simultaneously
the

intrinsic

feeling

the

course,

from

random

rule

of posing Bayes

no status to a

one

P(h,

a mathematical
one

In particular,

P(h, d)s

variables

on

insists

distribution

but it is only

of

experimental

I
fear the wrath

I-know-not-what
be claimedI

PSQM

mathematical

and

coexist. As always, hand d stand only for

consequences

Of

joint

meaning.

nature; without the intervention,

38

way

it,I
give

P(h, d). If

P(d)P(h|d)

one can do so of course,

without

Vd

simply

way. Instating

the product

artifice
P(h,d),

set

operator

unitary

a distinction

probability

convex

the

the final conditionalization

is

rule and the usual

joint

outside

falls

not understand

my

will bring down

can see

it

interventions

there is no hand

into

no d.

choice of words any

upon me. For

now, rather

it will

violentlythat

I
do

the first thing of what the problem

quantum measurement is:It is to supply

a mechanism

of
for

understanding how collapse

But

my

to leave

nothing

emphasized
taskand

intrinsic

comes

about, not to dismiss

it

language is honest language and meant explicitly

hidden.

The

cannot

use

a piece

here,

is that

be the taskof

to justify

how
that

as

already

it is not

theory

of information

of probability

gotten hold of

point

case,

in the classical

that

an agent
causes

the

makes

has

him to

change his beliefs. A belief is a property of ones head, not

of the object of ones interest.

38
anything that

can cause an application

rule within PSQM

But if there is nothing

applications

of

Bayes

more

rule

of Bayes

than arbitrary

to

ground

the

concept of quantum measurement, would not the

solidity of quantum theory melt away? What
else

can

determine when this rather than that

measurement

is performed?

Surely

that

much

has to be objective about the theory?

What Else Is Information?

Thats

territory

Im

not yet ready

to follow

you

into.

Good luck!

N.David Mermin, 2002

Suppose

one wants to hold

adamantly to the

idea that the quantum state is purely subjective.

That

is, that

there

is

no

right

quantum state for a systemthe

is nu-merically

additional

and

true

quantum state

to the

quantum

system. It walks through the door when the

agent

who

is

interested

in the system

walks

one consistently uphold

this point of view at the same time as supposing
that which POVM {Ed } and
which time
state-change
th esame
a ssuppo
d
rule d = Ad A
a measurement device
performs are objective features of the device?
The answer is no, and it is not difficult to see
through the door. Can

why.

Take
performs

{d }

as an example, a device that supposedly

a standard von Neumann measurement

the measurement

of which is accompanied

by the standard collapse postulate. Then when

click d is found, the posterior quantum state will

be d

= d

regardless

of the initial state . If

this state-change
the

or

device

systemi.e.,

rule is
its

it has

state d too must be

feature of

with

the

to do with the

nothing

subjective

observers

an objective

interaction

judgmentthen

an objective

the final

feature of the

quantum system. The argument is that simple.

Furthermore,

change

it clearly

operators

without

to all state

generalizes

rules for which

are

the Ad

any

adding

rank-one

further

complications.

Also though, since the operators

through

Ad

even the Ed

39

support

the maximal

=Ud

themselves

For otherwise,

one

Ed

control

of the final state

1/2

are

it must be that

subjective judgments.

would have

a statement

like,

Only states with support

would have
within
a statement
a subspacelikeSd

are correct.

Allother states

are simply wrong.

40

now of uninterrupted quantum time

as the special case of what happens to a

Thinking
evolution

state

after

the

single-element

POVM

{I}

is

performed, stat
one e
isafte
forced
rthe
to the
single
same
-element
conclusion

even

in

that

super-operator

case. The time

a quantum

for

evolution

systemmost

a completely positive trace-preserving

map on the space of operators for HD [50]
is a subjective judgment on exactly the same
par as the subjectivity of the quantum state.
Here is another way of seeing the same
generally

linear

thing. Recall what

argument

powerful

I
viewed

subjectivitythe

to be the most

quantum

the

for

Section 3. Since

we can

states

argument

Einsteinian

toggle

of

the quantum

a distance, it must not be something

over there, but rather something sitting
over here: It can only be our information about
the far-away
system. Let
us now apply a
state from
sitting

variation of this argument to time evolutions.

Consider

simple

quantum

bipartite

quantum

system

controlled

unitary operation Ui

us say

(For simplicity, let

on a
a

circuit

that

performs

on the target

bit.

the bipartite system

consists of two qubits.) Which unitary operation

the circuit two

applies
depends
upon
which
state |ii,
qubits.)
Which
unitary
operation
t
i 0,1, of

=
39

The support

of

an operator

by its eigenvectors with

40

Such

a statement,

nonzero

is the subspace

spanned

eigenvalues.

in fact, is not

so

dissimilar

to the

one
see

39

found in Ref. [95]. For several rebuttals

Ref. [2] and [96].

of that idea,

Figure 3: One
Einsteins

can use a

argument

slight modification

for the subjectivity

quan-tum state to

draw the

quantum

evolutions.

time

same

conclusion

By

of

of the
for

performing

on a far away system, one will

one or another completely positive map

measurements
ascribe

to

the

Therefore,

evolution

accepting

of

the

left-most

qubit.

physical locality, the time

evolution

map so ascribed cannot

be

a property

intrinsic to the system.

upon the control

an arbitrary state |i on the target,

two orthogonal states impinges

bit. ortho
two
Thus, for

one finds
|ii|i |ii(Ui |i)
(113)
for the overall evolution. Consequently

the

evolution of the target system alone is given by

|i Ui |i

(114)
On the other hand,

prepared

+ |1i.

in

suppose

superposition

the control bit is

state |i

a completely

positive

map

is,

||

|0i

Then the evolution for the target bit will

be given by

|i

U1|ih|U

Now,

(|ih|)

= ||

That

U0 |ih|U

(115)

to the point. Suppose

rather

than

feeding

feed half

we

single qubit into the control bit,

of

an

entangled

pair, where the other

qubit is physically far removed from the circuit. If

an

observer

with this description

of the whole

a measurement on

set-up happens to make

the

any
maps on
upon which

far-away qubit, then he will be able to induce

a number of completely positive

the control bit. These will depend
of

measurement he performs and which outcome

he gets. The point
is the same as before:
Invoking physical locality,
time

evolution

mapping

one obtains that the

on the single qubit

cannot be an objective state of affairs localized

at that qubit. The time evolution, like the state,
4142

is subjective information.
41

Of

try

course,

to get

could

argue

are

there

around

evolution

specified

opinion,

an effective

for the system

by the complete
however,

the Everettic
Einsteinian

conclusion.

For

that, The time evolution

control qubit is only

true

moves one can use to

sideways

this

moves

temple.

argument

toggles nothing

evolution

like this

One

in the

could

my

just prostrations

to

dismiss

same way:

with his localized

evolution

[97] In

circuit.

are

on the

for it.The

is the unitary

quantum

one

instance,

operator

the original

The

measurement;

observer
the true

quantum state is the universal quantum state. All that is

going

on in a quantum measurement

is the revelation

of

relative

statei.e.

can one argue

the effective

quantum

with this, other than to

most productive

stance

say

state.

How

it is not the

and that the evidence

shows that

since 1957 it has not been able to quell the foundations

debate. See Footnote 12.

42

the

of this argument

A strengthening

same way as

may

also be found in

in Section 3: Namely, by considering

the

teleportation of quantum dynamics. I

will for the moment,
however, leave that

many

as an exercise

for the reader. See the

references inRef. [98] for appropriate background.

40
It has long been known
preserving

over a
in a

completely

one-to-one

on a

the trace

linear maps
space can be placed

positive

D-dimensional vector

operators

that

correspondence
2-dimensional

with

space

density
via the

relation[79, 99, 100]

=I
(|ME

ihME

(116)

where |ME i
signifies

state

on HD

HD

|)

a maximally

entangled

|ME

Xi=1

|ii|ii

(117)
This

is

usually

representation

treated

as a

convenient

theorem only, but maybe it is

mathematical

accident

no

Perhaps there is a deep

reason for it:The time evolution one

to a quantum system IS a density
operator! It is a quantum state of belief no
more and no less than the initial quantum state
one assigns to that same system.
How to think about this? Let us go back to
the issue that closed the last section. How can
one possibly identify the meaning of a
measurement
in the Bayesian view, where a
physical

ascribes

measurement

ascription is itself subjectivei.e.,

a measurement

finds

a mathematical expression
of some agents
difficulty. When one agent
a piece of data d,he might

only inthe subjective refinement

beliefs? Here is the

contemplates

viewing

be willing to

use

the data to refine his beliefs

according to

P(h)

Xd
X

P(d)P(h|d)

(118)

However there is nothing to stop another agent

from thinking

the

same

Q(h)

data warrants

refine his beliefs according to

Xd

Q(d)Q(h|d)

him to

(119)
A priori, there need be

no

relation between the

Ps and the Qs.

A relation

only

comes

one

when

criterion for when the two agents will

they believe they

are drawing

the

same

seeks

say

that

meaning

from the data they obtain. That identification

is a purely voluntary act; for there is no

way

for

the agent to walk outside of his beliefs and

see

the world

as

it completely

and totally is.

The

standard Bayesian solution to the problem is

this:

When

statistical

both

model

agents

accept

the

for their expectations

same
of the

data d given

to the identity

say

they

considering.

are

I.e.

performing

measurement when and only when

P(d|h)

,
.

of the measurements

each (separately)

will

agree
they are

hypothesis h,then they will

= Q(d|h)

(120)

two agents

the

same

h and d

Putting this in a

more evocative form, we can

say that both agents agree to the meaning of a
measurement
when
they
adopt
the
same
resolution of the identity

Xd

P(d)P(h|d)

P(h)

Xd

Q(d)Q(h|d)
Q(h)

(121)
with which to describe it.

With this, the relation to quantum

measurement should be apparent. If we take it

a measurement is anything that

a refinement of ones beliefs, then an
agent specifies a measurement when he specifies a
seriously

that

generates

resolution of his initial density operator

Xd

P(d) d

(122)

41
But again, there is nothing to stop another

agent from thinking the data warrants

refinement

that is completely unrelated to the

first:

Xd
X

Q(d) d

(123)

And that is where the issue ends if the agents

have

no further agreement.
as inthe classical case, however,

Just

there

is a solution for the identification problem.

Using the canonical construction of Eq.(86), we

can say that both agents agree to the meaning
of a measurement when they adopt the same

resolution of the identity,

Xd

P(d)

1/2

1/2

Xd

Xd

Q(d)

1/2

1/2

(124)

with which to describe it.

Saying it ina more tautological

way, two
on the identity of
assign it the same

agents will be inagreement

a measurement

when they

POVM {Ed },
Ed

= P(d) 1/2 d

Q(d)

1/2

1/2

1/2

=
(125)

move, however, is that

out the proper way to think about the

The importance
it draws

operators

Ed

of this

from

the

present

perspective.

They play part of the role of the statistical

model

P(d|h).

More

generally,

that

role

fulfilled by the complete state change rule:

P(d|h)

(126)
That is to say, drawing the correspondence in
different terms,

is

P(d|h)

= Ud

d ()

1/2

1/2

Ed

(127)
(Of

course,

more

measurementsd

sum of

generallyfor

may

()

consist

nonefficient
of

a convex

such terms.)

The completely

mathematical

positive

description

a map.

evolution is just such

map
to

that

gives

quantum

time

Its role is that of the

subjectiveis just
statistical
such amodel
map. Its
P(d|h),
role is
where
that of
d just
happens

to be drawn from a one-element set.

Thus, thinking back

on entanglement,

it

seems

the general structure of quantum time evolutions

cannot
What

the wedge

we see

we are

instead

looking

for either.

is that there is

a secret

waiting to be unlocked, and when it is unlocked,

it

will

very

likely

quantum measurements.

Intermission

us as much about
as quantum states and

tell

quantum time evolutions

us take a deep breath. Up until now I

have
as much quantum mechanics
as I
could, and I
know that takes a tollit has
taken one on me. Section 3 argued
that
Let

tried to trash about

quantum

stateswhatever

they arecannot

be

objective entities. Section 4 argued that there

is nothing

sacred about the quantum probability

rule and that the best

state is

as a state
if one were

way to think

of

a quantum

of belief about what would

a standard
away in a vault in
Paris. Section 5 argued that even our hallowed
quantum
entanglement
is
a secondary and
subjective effect. Section 6 argued that all a
measurement is is just an arbitrary application of
happen

measurement

Bayes

to

device

rulean

beliefsalong

measurements

ever

arbitrary
with

are

approach

locked

refinement

some

invasive

account

of ones

that

interventions

nature. Section 7 argued that

into

even quantum

time

42
evolutions

are

subjective judgments; they just

so

happen to be conditional judgments.

it went.

...

And,

It is

Subjective. Subjective! Subjective!!

word that will not

not something
sake. There

is

go away.

But subjectivity

to be worshipped

are

so

a
is

for its

own

we

need

limits: The last thing

bloodbath of deconstruction. At the end of

some term, some

the day, there had better be

element

in quantum theory that stands for the

objective,

or we might as well melt away

and call

this all a dream.

I
turn now to a more constructive phase.

Unknown Quantum States?

My thesis, paradoxically, and

a little

provocatively, but

nonetheless genuinely, is simply this:

QUANTUM STATES DO NOT EXIST.

The abandonment

. .,

of superstitious

istence of Phlogiston, the Cosmic

and Time,

or Fairies

beliefs about the

Ether, Absolute

and Witches,

ex-

Space

was an essential step

along the road to scientific thinking. The quantum state,

too, if regarded

as something

objective existence, is

no

less

illusory attempt to exteriorize

endowed with

a misleading

some

kind of

conception,

or materialize our true

an

prob-

abilistic beliefs.

the true ghost of Bruno de Finetti

The hint of

a more

fruitful direction

found by trying to make

most

commonly

information

theory

phrases
from

is

hardly

information

that

a paper
not

does

Unknown

quantum

protected

with quantum

[101],

be

of the

quantum

Bayesian

and

used

to

But

what

information-based

can

check

the

use

make

are
error

states

quantum

in

of it

teleported

[23]

correcting

codes

quantum

for

eavesdropping [102]. The list of

day.

in

can

one

of

It is the unknown quantum state

perspective.
There

used

sense

uses grows

term mean?

interpretation

of

each

In

an

quantum

an oxymoron: If quantum states,

very definition, are states of subjective

mechanics, it is
by their

information

and not states of nature, then the

state is known by someoneat

the

person who wrote it down.

Thus, if a phenomenon ostensibly
concept of an unknown state in its

very

least,

by the

invokes the
formulation,

that unknown state had better be shorthand for

more
for

basic situation (even if that basic situation

a complete

still awaits

any

analysis)

quantum

state

in its

means that
an unknown
we should

This

phenomenon using the idea of

description,

demand that either

owner of the unknown statea further

decision-making agent or observerbe explicitly
1. The

case, the unknown state

a stand-in for the unknown state of
of an essential
player
who went

identified. (In this

is merely
belief

unrecognized inthe original formulation.) Or,

no

further agent

or

then

a way must

be

2. If there is clearly
observer

on

the

scene,

found

to

reexpress

the

phenomenon

with the

term unknown state completely banished from

its
formulation.

(In this

of the effort will be

case,

the end-product

single quantum state used

for
describing

state
overall

that

the phenomenonnamely,

actually

captures

the

the

describers

set of beliefs throughout.)

This Section reports

and [33], where such

the

experimental

tomography

[31]

tomography is

the work of Ref. [32]

a project

practice

The

is carried out for

of

usual

quantum-state
description

of

43

some sort, say a nonlinear

by a laser, repeatedly
prepares many instances of a quantum system,
say many temporally distinct modes of the
electromagnetic
field, in a fixed quantum state ,
pure or mixed [103]
An experimentalist
who
this. A device

optical medium

of

driven

wishes to characterize

or to

the operation of the device

use may be able to

on the systems it

calibrate it for future

perform

measurements

prepares even if he cannot get at the device

itself. This can be useful if the experimenter
has some prior knowledge
of the devices
operation
that can be translated
into a
probability
distribution
over states. Then
learning

about the state will also be learning

about the device. Most importantly, though, this

description
precise

tomography

of

state

experimenter

is unknown.
is

measurements,

to

(on

enough

that

the

The goal of the

enough

perform

and

measurements

assumes

kinds

of

to

large enough sample),

estimate the identity of .

This is clearly

an example

where there is

no

player on whom to pin the unknown

as a state of belief or judgment. Any
attempt to find such a missing player would be
further

state

entirely

artificial:

Where

would

the

inside

of

placed?

On

tomographer
only

device

be

the

is trying

to characterize?

43

course

is the second

strategy

available

aboveto

the player

the

The

banish the idea of the unknown state

from the formulation of tomography.

To do this, we

once

Bayesian probability
phasized

previously,

probabilitiesjust

objective
of

again take

our cue from

em-

theory[75, 76, 77]. As

in

Bayesian

states of nature, but rather

belief,

theory

like quantum statesare

reflecting

ones

not

measures

operational

in vari-ous gambling scenarios. In

commitments

light of this, it

most

the

comes as no surprise

overarching

Bayesian

that

one

of

is to

themes

a set of
decision-making agents can come to a common
belief or probability assignment for a random
variable even though their initial beliefs may
the conditions

identify

differ[77]. Following

understanding

the

under

which

that theme is the key to

essence

of

Bayesian

theory

quantum-state

tomography.
Indeed, classical
almost

precisely

same

of the phrase unknown

is

This

egregious

as unknown state.

The procedure
tomography

an

this

analogous

every

in its
bit

as

to quantum-state

in Bayesian theory is the estimation

unknown probability

repeated trials

The

as our

probability

an oxymoron

domain.

of

encounters

problem

quantum state through the widespread

unknown

use

the

on identically

way to eliminate
situation was

from the results of

prepared systems.

unknown probabilities

de

Finetti in the early 1930s [104]. His method

was

to focus

on

the

by

from

Bruno

simply

introduced

equivalence

of

the

repeated

trials

important

is

that

are

any

that

as

far

concerned.

probability

what

assignment

is really

are

systems

the

as

indistinguishable
predictions

namely,

probabilistic

, ,...,

Because

p(x1

x2

xN

for multiple

trials should

permutation

of the systems. As innocent

conceptual shift

use

it

may

representation
multi-trial

effect.

theorem,

probability

permutation-symmetric
number

of

multi-trial

equivalent

be symmetric

sound, de Finetti

to powerful

trialsde

he

an

probability

was

with

showed

as

this

able to
his

any

that

called

under

that

arbitrarily

Finetti

probabilities

to

For,

assignment

for

this,

of

is

large
such

exchangeableis

for the unknown

probabilities.

Let us outline this in a little more detail. In

an objectivist description of N identically prepared
systems,
the
individual
trials
are
described by discrete random variables xn {1,
2,
k}, n = 1,
N,and the probability in
the multi-trial hypothesis space is given by an

..., ...,

independent

43

Placing

the

as

respectable

anonymous

here

would

Berkeleys

be

famous

about

as

patch to his

system of idealism. The difficulty is captured

philosophical
engagingly

player

George

by

limerick

Ronald

of

Knox

and

its

reply:

There

was a young man

think it exceedingly
Continues

odd

to be

who said, God

:
If he finds

:
When

:
Must
:

that this tree

no one

theres

about in

the Quad. REPLY: Dear Sir: Your astonishments

odd.

:
I
am

thats why the tree

always about in the Quad.

:
Will continue

:
And

to be, :
Since

observed by Yours faithfully, God.

44
identically distributed distribution
p(x1,x2

=p

n1
1

p2

The numbers

nk
k

=px1

,...,xN

n2

px2 pxN

(128)

,..., .

pj

describe

that the result of

will be j (j

=1

other

hand,

k)

describes

the objective,

probability

a single

The variable nj

the

number

true

experiment

of

on the
times

outcome j is listed inthe vector (x1

, ,...,
x2

xN

But this descriptionfor

the objectivistonly

describes the situation from

a kind

,...,

of Gods eye

point of view. To the experimentalist, the true

probabilities

p1

pk

will

very

often be

unknown at the outset. Thus, his burden is to

estimate

the

unknown

probabilities

statistical analysis of the experiments

by

outcomes.

In the Bayesian approach, however, it does

sense to talk about estimating a true

a Bayesian assigns a prior
probability distribution p(x1 x2
xN )on the
multi-trial hypothesis
space and uses Bayes
not make

, ,...,

probability. Instead,

theorem to update the distribution

measurement

of his

in the light

results. The content

of de

Finettis theorem is this. Assuming only that

p(x(1)

,x(2) ,...,x(N)

= p(x1

,x2 ,...,xN

(129)
for

any

permutation

that for
pN+M

any

(x1,x2

of the set {1,...,N},

and

integer M > 0, there is a distribution

,...,xN+M

)with the

permutation property such that

same

p(x1,x2

,...,xN

pN+M (x1,...,xN ,xN+1 ,...,xN+M

X

(130)
xN+1 ,...,xN+M

then p(x1 ,x2

,...,xN

)can be written uniquely in

the form
p(x1,x2

,...,xN

ZSk

P(~ p)px1 px2

pxN

d~ p

ZSk

n1

n2

P(~ p)p 1 p 2

nk
k

d~ p

(131)

p = (p1,p2 ,...,pk ),and the integral

taken over the simplex of such distributions

where

p:
pj

Sk

0 for all j and

pj
j=1

(132)

=1

is

Furthermore, the function P(~ p) 0 is required to

be

a probability

ZSk

P(~ p)d~ p

density function

on the simplex:

=1

(133)

With

this

representation

concept

unsatisfactory

probability

vanishes

an

from the

favor of the fundamental

exchangeable

theorem,

of

the

unknown

description

in

idea of assigning

an

distribution to multiple

probability

trials.
With this

to

reword

cue

the

in hand, it is
description

tomog-raphy to meet
is

simply

experimenter

our

judgment
notice

easy to see

of

quantum-state

goals. What is relevant

on

the

part

the essential

character of this judgmentthat

the systems

distinction

between

preparing.

In operational

judgment that all the systems

same as

far

as

how

the

of

subjective

there is

the device

terms,

this

is

no

is
the

are and will be the

predictions are

observational

concerned. At first glance this statement

might

seem to be

contentless, but the important point

is this: To make this statement,

use

the

notion

of

an

one

need

never

statea

unknown

completely operational description is good

45
enough.

Putting

statement

it into technical

terms, the

is that if the experimenter

judges

collec-tion of N of the devices outputs to have

an overall quantum state (N), he will also

judge

any

permutation

outputs to have the

of

those

same quantum state

(N).

no matter

how large the

number N is. This, complemented

only by the

Moreover, he will do this

consistency

(N)

condition that for any N the state

(N+1),
from
makes for the

be derivable

complete story.

The words quantum

formulation, just
of

tomography,

as
but

state

appear

in this

in the original formulation

there

is

no

longer

any

mention of unknown quantum states. The state

(N)

is known by the experimenter

else), for it represents

(if

his judgment

no one

More

importantly,

the experimenter

an

make

structure of the
Each

has

the

of

whole

states

is in a position to

statement

unambiguous

sequence

about

of states

the
(N):

(N)

kind of permutation

invariance

over

its

factors. The content of the quantum de Finetti

theorem[32,
105]
is that
a
sequence of states (N) can have these properties,
which are said to make it an exchangeable
sequence, if and only if each term init can also be

representation

written inthe form

ZDD

(N)

P()

(134)
where

is

an N-fold

tensor product. Here P() 0 is a fixed

probability

space
ZDD

DD

distribution

and

P()d

=1

over

the density operator

(135)
where d is a suitable
The interpretive

measure.
import

of this theorem

paramount. For it alone gives

is

mandate to the

term unknown state in the usual description of

tomography.

act

as if

It

says

his judgment

he knows there is

can

that the experimenter

comes

(N)

a man

about because

in the box, hidden

from view, repeatedly preparing the same state .

He does not know which such state, and the

best he

can say

about

the unknown

state is

captured inthe probability distribution P().

The quantum de Finetti theorem furthermore

makes

a connection to the

Bayesianism

stressed

two independent

overarching

theme of

above. It guarantees

observersas

long

as they

for
have

rather minimal agreement in their initial beliefs

that the outcomes of

set of measurements

sufficiently

will force

informative

a convergence

in

their state assignments for the remaining systems

[33]

by

This minimal

agreement is characterized

a judgment on the part

of both parties that

sequence

the

of

described

above,

observers

are not

systems
and

though P()

can never

may

promise

absolutely

opinions. Quantitatively,

the latter

the

in their

means that
zero, it

be arbitrarily close to

vanish.

exchangeable

rule

gathered

quantum

by another
for

updating

an
be

on K systems

from

version

probabilities

of

[33]

yield

then the state of additional systems is

probability

as in Eq. (134), but

on density operators

P(|DK )

constructed

sequence can

information

Specifically, if measurements

results DK

works because

density operator

to reflect

measurements
Bayess

that

inflexible

This coming to agreement

updated

as

is exchangeable,

P(DK |)P()
P(DK )

(136)

using

Here P(DK |) is the probability

measurement

results

DK

an updated

given by

to obtain the

given the state

for the Kmeasured systems, and

ZDD

P(DK )

P(DK |)P()d

(137)

46
is

the

unconditional

measurement
informative

probability

a
measurements, as

results.

set of

For

large, the updated probability

highly peaked

on a

for

K becomes

P(|DK )becomes

particular state DK

dictated

by the measurement

results, regardless

prior probability P(),

as long as P()

in

neighborhood

observers

have

encapsulated

of DK
different

in different

the

sufficiently

the two

Suppose

initial

of the

is nonzero

beliefs,

= 1,2.

priors Pi (), i

The measurement results force them to a common

state

of

belief

in which

additional systems
N

DK

number

assigned the product

i.e.,

Pi (|DK

(138)

are

any

DK

N of

state

independent

of i,for K sufficiently large.

on the purpose

This shifts the perspective

quantum-state

tomography:

uncov-ering some
rather

about

the various

over

agreement

unknown

It

quantum

the

observers

coming

future probabilistic

de

about

state of nature, but

In this connection, it is interesting

conclusions

not

is

of

Finetti

to

predictions.

to note that

theorem

and

the

just drawn from it work only within

mechanics.
real Hilbert

com-plex vector-space quantum

mechanics based on
spaces [106] the connection between

the framework

of

For quantum

exchangeable

operators

density

and unknown

quantum states does not hold.

A simple

counterexample

is the following.

Consider the N-system state

(N)

1
2

12

(139)
where
+

1
(I+ 2 )
2

and

12
(I 2 )

(140)

and 1

,,
2

are

and 3

(139)

a valid
to an equally

is clearly

corresponds

the Pauli matrices. In

quantum

complex-Hilbert-space

mechanics,

operator:

density

Eq.

It

weighted mixture of

Nspin-up particles and Nspin-down particles in

the

y-direction.

state

The

(N)

is

thus

exchangeable, and the decomposition inEq. (139)

to the quantum de Finetti

is unique according
theorem.

But

now

consider

real-Hilbert-space
its ostensible
remains

because,

(N)

as an operator

quantum mechanics. Despite

use of the imaginary

a valid quantum
upon expanding the

Eq. (139), all the terms with

2 s cancel

away.

exchangeable

Yet,

density

written in de Finetti

, ,

real symmetric
i2

Finetti

even

number i,it

state.

This

right-hand

an odd
it

side of

number of

though

operator,

is

it is

cannot

an

be

form Eq. (134) using only

operators. This follows because

as a linear combination of
and 3
while a real-Hilbert-space de
expansion as in Eq. (134) can only

cannot be written

I, 1

in

contain

those

Finetti

theorem

operators.

three

does

not

Hence

hold

the de

in

real-Hilbert-space quantum mechanics.

In

classical

exchangeability

probability

characterizes

where the only data relevant

probability
i.e.,

distribution

the

numbers

are

nj in

theory,
those

same

true

is

(131).

to

robust measurement, in particular,

is informationally
updating

where

an

more can

sequential measurements.

mechanics:

a sufficiently
any one that
sufficient

state to the

exchangeable

nothing

are

complete)

be

The

shows that

in quantum

data (with respect

Frequency

data,

frequency

Eq.

quantum de Finetti representation

the

situations

for updating

learned

That is,

one

for

point

from

obtains

a convergence of the form Eq. (138), so that

ultimately
any further measurements on the
individual

systems

independent.

That

Finetti theorem

that there

are

will

in real

be

statistically

no quantum de
Hilbert space means

there

is

fundamental differences

real and complex Hilbert

spaces

between

with respect to

learning from measurement results.

Finally, in summary, let

of learning for just

a little

us hang on the point

longer. The quantum

essence

de Finetti theorem shows that the

quantum-state

tomography

of

is not in revealing

an
47
element of reality but inderiving that various

agree some minimal amount) can

come to agreement intheir ultimate
agents (who

quantum-state assignments. This is not at all the

same

as the statement reality does not

It is simply that one need not go to the

thing

exist.

extreme of taking the unknown quantum state

as being objectively

real to make

sense

of the

experimental practice of tomography.

J.M.Bernardo and A.F.M. Smith intheir

book Ref. [77] word the goal of these exercises

we

have

run through

inthis paper

very

nicely:

[I]ndividual degrees of belief, expressed

as probabilities, are inescapably

the starting

point for descriptions of uncertainty.

There

can be no theories

without theoreticians;

no learning

no science

without learners; ingeneral,

without scientists. It follows that

processes, whatever their

concerns and fashions at any given

learning
particular

point in
time, are necessarily reasoning

processes

which take place inthe minds of individuals. To

be sure, the object of attention and interest

may

well be an assumed external, objective

reality: but the actuality of the learning

process

...
..

consists inthe evolution of individual,

subjective beliefs about that reality

However, it is important to emphasize

that

the primitive and fundamental notions of

individual preference and belief will typically

provide the starting point for interpersonal

communication and reporting

processes.

[W]e shall therefore often be concerned to

identify and examine features of the individual
learning

process

which relate to

interpersonal issues, such as the conditions under

which

occur

an approximate consensus of beliefs

in a population of individuals. The

might

acase in
a population can

quantum de Finetti theorem provides

point for how much agreement

come to from within quantum mechanics.

One is left with a feelingan almost

salty

feelingthat perhaps this is the whole point of

the structure of quantum mechanics. Perhaps
the missing ingredient for narrowing the

structure of Bayesian probability down to

quantum mechanics has been infront of us all

along. It finds

no

better expression than in

taking account of the challenges the physical

world

poses to our coming to agreement.

10

The Oyster and the

Quantum

The
insight

significance

of

into the logical

this

development

possibility

of

is to give

a new

us

and wider

pattern of thought. This takes into account the observer,

including

way

it

was

of thought

observer,

the apparatus

we

do not

oc-curring

type of theory, but

...

used by him, differently

done in classical physics

assume any

In the

longer

in the idealizations

an observer

from the

new pattern

the detached

of this

classical

who by his indeterminable

a new

effects creates

new state

observation is

system.

a singling out

now,

here and

...

situation, theoretically

the observed

of

of

a particular

from the theoretical

making obvious

described

Inthis

factual result,

possibilities,

aspect

the discontinuous

as a

way every

thereby

of the physical

phenomena.
Nevertheless,
the-ory

any

deny
of

an

there

objective

remains

reality,

as

kind of
theories

for the observer to influence the results

possibility

a measurement, once

cho-sen. Particular

new

these

still in the

inasmuch

the experimental

qualities

of

an

arrangement

individual

observer

is

do

not enter the conceptual framework of the theory.

Wolfgang Pauli, 1954

48
A grain of sand falls into the shell of an oyster
and the result is

a pearl. The oysters sensitivity

source a beautiful gem. In

to the touch is the

the

75

founding

years

that

have turned to

may

have

passed

since

the

of quantum mechanics, only the last 10

view and

finally reveal the

essence

an

attitude

that

of the theory. The

quantum world is sensitive to the touch, and

that

may

Quantum

be

one

of the best things about it.

informationwith

specializations

of quantum

its

information

three

theory,

quantum

cryptography,

computingleads

that idea.

quantify

way

the

quantum

and

us how to
algorithms can

in telling

Quantum

be exponentially faster than classical algorithms

keys

Secret

systems

can

in such

be

encoded

into

a way as to

physical

reveal whether

information has been gathered about them. The

list of triumphs keeps growing.

The key to
in

change

so much

back to almost

going

of this has been simply

of attitude. This

any

can

older

be

seen

textbook

by

on

quantum mechanics: Nine times out of ten, the

Heisenberg

way

that

uncertainty

conveys

short-changed

relation is presented in

the feeling that

we

by the physical world.

Look at classical physics, how nice it is:We can

measure a particles position and momentum with as

much accuracy as we would like. How limiting
quantum theory is instead. We are stuck with

1
h
2

xp

and there is nothing

we can do about it.The task

up to this state of affairs and

of physics is to sober

make the best of it.

have been

How this contrasts with the point of departure

of quantum information!

There the task is not

to ask what limits quantum mechanics

places

upon us, but what novel, productive things we

can do in the quantum world that we could not
have done otherwise. In what ways is the
quantum

world

fantastically

better

than

the

classical one?
If

one

is looking for something

real

in

more direct tack could

one take than to look to its technologies? People
may argue about the objective reality of the
wave function ad infinitum, but few would argue
about the existence of quantum cryptography as
a solid prediction of the theory. Why not take
that or a similar effect as the grounding for what
quantum mechanics is trying to tell us about
quantum theory, what

nature?

us try to give this imprecise set of

some shape by reexpressing quantum
cryptog-raphy in the language built up in the
Let

thoughts

previous sections. For quantum key distribution

to be able to

it is essential

system in one

some

from

or

prepare a

set [102, 107]

nonorthogonal

fixed

physical

another quantum state drawn

are

These nonorthogonal states

used to encode

potentially secret cryptographic

key to be shared

between

receiver.

the

information

sender

an

and

eavesdrop- per

was

which quantum state

nonorthogonal

What is novel here

of the proposed

states

key into

the

forces

process

information-gathering

disturbance

is about

actually prepared in

each individual transmission.

is that the encoding

seeks

The

to

induce

to the overall set of states. That is,

pres-ence of an active eavesdropper transforms

pure states into a set of mixed states
or, at the very least, into a set of pure states
the

the initial

with larger overlaps

than before. This action

ultimately boils down to

for

the

receivers

sender

over

a loss

the

measurements

of predictability

outcomes

and,

so,

is

detectable by the receiver (who reveals

those

outcomes

for the

senders

More importantly, there is

the

some

inspection)

a direct

between the statistical information

of

directly

of

connection

gained by

an

and the consequent disturbance she

eavesdropper

process.
goes up, the
goes up in a precisely

must induce to the quantum states inthe

As

the

gathered

information

necessary

disturbance

also

formalizable way[108, 109].

Note the two ingredients

this scenario.

appear

that

or measurement

is grounded with respect to

case,

observer (in this

the eavesdropper)

49

is grounded with respect to another

(here, the sender)

is

In particular, the disturbance

to

disturbance

to

predict

measurements

the

perform. No hint of

system is made

use

idea of measurement
The
consider

second

at

preparations

the

senders

previous

is measured by her diminished

informationthis

ability

one

while

the

disturbance

in

First, the information gathering

outcomes

the

legitimate

any

certain

might

variable intrinsic to the

of in this formulation of the

causing disturbance.

ingredient

least

of

receiver

two

is

possible

that

one must

nonorthogonal

inorder for the formulation to have

any

meaning. This is because the information

gathering

is

variable

some
some

not about

observablei.e.,

about

or reality

classically-defined

unknown

hidden

intrinsic to the systembut

is

instead about which of the unknown states the

sender

actually

prepared.

The lesson

is this:

Forget about the unknown preparation, and the

random outcome of the quantum measurement

about

information
quantum

noise

nothing.

no

with

It

is

to

connection

is

simply

any

preexisting variable.
How crucial is this second ingredientthat
is, that

be at least

there

two nonorthogonal

states within the set under consideration? We

address

its necessity

account

above:

One

might

goal is not

eavesdroppers

a
say

by making

can

shift in the

that

the

so much to uncover

the

un-known quantum state, but to

predictability
over the receivers

identity of the
sharpen

her

measurement outcomes. In fact, she would like to

do this at the

senders
Changing

same

predictions
the

time

as

language

as

little
still

disturbing

as
further

the

possible.

to

the

terminology
actions
the

of

4, the

Section

serve to sharpen her information about

consequences of the receivers
interventions on the system. (Again, she

potential

further
would

to

like

diminishing
about

do

the

this

senders

same

those

cryptographic

context

more

point

minimally

previous

information

In

ultimately

the secret key. From

view,

of

comes to a

however,

importance of this change of language

leads to

an

As previously
probability
identify

the

decision-making

agents

assignment

the

within Bayesian

significant

theme

is to

a set of
a common

the conditions

probability

on

problem.

emphasized,

most

the

is that it

almost Bayesian perspective

informationdisturbance

the

byproduct of this effort

sound prediction of

present

while

consequences.)

is that the eavesdropper

the

eavesdroppers

under which

can come to
some random

for

variable

inspite of the fact that their initial probabilities

differ [77] One might similarly view the process of

quantum

eavesdropping.

eavesdropper

start

The

sender

off initially

and

the

with differing

quantum state assignments for a single physical

system. In this

can

sender

case

so

happens

that the

make sharper predictions

than the

eavesdropper

it

about

outcomes

the

of

receivers

measurements. The eavesdropper,

satisfied

with

measurement

this

on

situation,

the system in

performs
a
an attempt to

sharpen those predictions. In particular,

an attempt

to

come

into

the

not

there is

something

of

an

agreement with the sender but without revealing

the outcomes
her very

It is at this point that

of

quantum

the

The

world

eavesdroppers

come

or,

of her measurements

indeed,

presence.

into

distinct property

makes

attempt

alignment

with

predictability

is always

shunted

goal.

shunting

of

This

predictability

is the subtle

itself

known.

to surreptitiously
the

away

various

manner
our

quantum world is sensitive to

senders

from its

observers

in which the

experimental

interventions.

our crucial hint! The

a distinction between

And maybe this is

wedge
Bayesian

that

drives

probability

theory

in

general

and

quantum

mechanics

in particular

is perhaps

nothing

more

than this Zing! of

a quantum
when an agent

system

that

is

manifested

interacts with it.It is this wild sensitivity to the

touch that keeps

ever

our

information and beliefs from

an alignment. The
our beliefs about the potential
consequences of our interventions on a system
can come into alignment is captured by the
mathematical structure of a pure quantum state
coming into too great of

most

|i.

Take

curves

for

all possible

information-disturbance

a quantum system,

tie them into

bundle, and that is the long-awaited

input

we

property, the

have been looking for from nature. Or,

at least, that is the speculation.

50
10.1

What

Give Us

we need

a Little

Reality

here is a little Realitty.

Herbert Bernstein,

circa 1997

In the previous version of this

paper [1] I
ended

the discussion just at this point with the following

words, Look at that bundle long and hard and

we might

just find that it stays together without

our tie. But I

imagine that wispy
was singularly unhelpful to anyone
who wanted to pursue the program further.
How might one hope to mathematize
the
the help of

command

bundle

curves

of all possible

for

a system?

If it

effort will have to end

single

real

information-disturbance

can be done at all, the

up depending upon a

parameterthe

dimension

of

the

Hilbert space. As a safety check, let

us ask ourselves right at the outset whether this
is a tenable possibility? Or will Hilbert-space
dimension go the wayside of subjectivity, just as
we saw so many of the other terms in the
theory go? I
think the answer will be in the

systems

negative:

to be

Hilbert-space

dimension

a stand-alone concept

with

will survive

no

need of

an

agent for its definition.

The simplest check perhaps

same
the

Einsteinian

test for it

is to

as we

pose

the

did first for

quantum state and then for quantum time

evolutions. Posit

a bipartite system

with Hilbert

spaces

HD1

and HD2

respectively)

(with dimensions D1 and D2

and imagine

an

initial quantum

state for that bipartite system. As argued too

many times already, the quantum state must be

a subjective component inthe theory because the
theory allows localized measurements on the D1
system to change the quantum state for the D2
system. In contrast, is there anything

one can

do at the D1 site to change the numerical value

of D2 ? It does not

appear so. Indeed,

the only

way to change

that number is to scrap the initial

supposition.

Thus,

every

right

to call

to that

extent,

the numbers

D1

one

has

and D2

potential elements of reality.

may not look like much, but it is a start. 44

And one should not belittle the power of a good
45
hint, no matter how small.
It

11

Appendix: Changes Made

Since quant-ph/0106166 Version

Beside overhauling the Introduction

so as to

make it more relevant to the present meeting, I

made the following

more

substantive changes to

the old version:

1. I
made the language

slightly less flowery

throughout.

are now

2. Some of the jokes

the

readers

who

thought

explained

they

for

were

typographical

errors.
3. For the
labeled

purpose

the

of Section 1s imagery, I

followers

of

the

Spontaneous

Collapse

and
Spontaneous

Many-Worlds
Collapseans

interpretations,
and

Everetticsin

contrast to
the

previous
and

Collapsicans

terms

Spontaneous

better

Everettistasto

emphasize their

religious aspects.

4. Some

figures

were

removed

from

the

quantum de Finetti section and the dramatis

personae
on page 2 was added.
5. Inow denote the outcomes

of

general

POVM by the index d to evoke the image that all

(and

a quantum measurement ever does

a piece of data by which we update our

only)

gather

subjective
It

probabilities

for something

causes us to change our subjective

44

45

is

else.

probability

Cf. also Ref. [110].

Cf. also the final paragraphs of Section 1.

51
assignments P(h) for

some

hypothesis h to

posterior assignment Pd (h) conditioned

on the

data d.

6.As noted in Footnote 9, this paper is a bit

a transitionary one for me inthat, since writing
quant-ph/0106166,

more

convinced

Ihave

of the consistency

become

of

much

and value of

the
radically

subjective

Bayesian paradigm for

probability theory. That is,I

have become much

more

inclined

Finetti [104]
[111]. To

say,

to the

view

of

Bruno

de

than that of Edwin Jaynes

that end, I
have stopped calling probability

distributions states of knowledge and been

more true to

the conception

that they

are

states of belief whose cash-value is determined

way an agent will gamble in light of

them. That is,a probability distribution, once it
is written down, is very literally a gambling
commitment the writer of it uses with respect to
by the

the phenomenon

he is describing. It is not

clear to what extent this adoption of terminology

cause obfuscation rather than clarity in

paper; it was certainly not needed
for many of the discussions. Still I
could not
stand to propagate my older view any further.
will

the present

7. In general, 23 footnotes, 38 equations, and

over 43 references have been added. There are

five new historical
quotes starting
the
sections, and the ghostly quote of Section 9 has
been

modified for greater

8. The metaphor

accuracy.

ending Section 1, describing

how the grail of the present quantum foundations

program can

be likened

to the spacetime

manifold of general relativity, is new.

9.Section 2 has been expanded to be consistent

with the rest of the paper. Also, there

are three

footnotes to be found

important explanatory
there.

10. Einsteins

letter to Michele Besso in Section

3 is now quoted infull.

argues more

11. Section 4.1, which

strongly

for Gleasons noncontextuality assumption than

previously, is new.

12. Section 4.2, which explains informationally

complete POVMs and
standard

uses

quantum

them to imagine

measurement

at the

Bureau of Weights and Measures, is new.

13.

To

elaborate

entanglement

and

the
the

connection
standard

between

probability

rule, I
switched

the order of presentation

of the

Whither Bayes Rule? and Wither Entangle-

ment? sections.

14. The technical mistake that

was

in Section 5

is now deleted. The upshot of the old argument,

however,

remains:

The tensor-product

for combining quantum systems

rule

can be thought

of

as

secondary

to the structure

of local

observables.

15.

much

classical

greater

measurement

elaboration

of

the

problemi.e.,

the

mystery of

physical cause of Bayesian conditionalization

upon the acquisition of new information (or the
lack of a mystery thereof)is now given in
Section 6.

16.

Section

description

6.1,

wherein

of the relation

a more
between

detailed
real-world

measurements

and

the

hypothetical

standard

quantum measurement is fleshed out, is new.

argues for the nonreality of

necessary subjectivity
of the ascription
of a POVM to a
measurement device, is new.
18. Section 8, wherein I
find a way to use the
word bloodbath, is new.
17. Section 7, which

the Hamiltonian and the

19. The long quote in Section 9 by Bernardo and

Smith, which describes what Bayesian probabil-

ity theory strives for, is

new. Heres

another

good quote of theirs that didnt fit inanywhere

else:

What is the nature and scope of Bayesian

Statistics within this spectrum of activity?

52

Bayesian Statistics offers

of personalistic

rationalist

beliefs in contexts

theory

of uncertainty,

with the central aim of characterising

kinds of undesirable
theory

decision

behavioural

establishes

maximization

provides

making

that
the

and

fit

together

inconsistencies.

expected

basis

that

provides the key to the

should

an

act in order to avoid certain

individual should

The

how

Bayes

ways

in the

utility
rational

for

theorem

in which

light

of

beliefs

changing

evidence. The goal, in effect, is to establish rules

and

procedures

for individuals

concerned

with

disciplined uncertainty

accounting. The theory is

not descriptive, in the

sense

of claiming to model

actual behaviour. Rather, it is prescriptive, in the

sense

you wish to avoid the possibility

undesirable consequences you must act

of saying if

of these

inthe following way.

20. Section 10.1, which

argues

for the

nonsubjectivity of Hilbert-space dimension, is

new.

21. One

can read about

the term Realitty

in Ref. [112].

12

Acknowledgments
I
thank

Carl Caves, Greg

corre-spondence

Asher

Cabello,

Plotnitsky
dramatis

years

of

that led to this view, Jeff Bub and

Lucien Hardy for giving

Ad an

David

Comer,

Mermin, and R udiger Schack for the

for

their

personae

of

me courage
Peres,

help
the

in general,

and

Arkady

in compiling
Introduction,

the
Jeff

Nicholson for composing the papers figures, and

Andrei Khrennikov

thanks
Steven

for infinite patience. Further

go to Charlie Bennett, Matthew

van Enk, Jerry Finkelstein,

Donald,
Philippe

Grangier, Osamu Hirota, Andrew Landahl, Hideo

Mabuchi,

Jim Malley,

Mike

Nielsen,

Masanao

Ozawa, John Preskill, Terry Rudolph, Johann

Summhammer,

for their

Chris Timpson,

and Alex Wilce

many comments on the previous

version

of this paperall of which I

tried to respond to in

some
out

or fashionand particularly warm

goes to Howard Barnum for pointing

shape

gratitude

my

technical

Entanglement?

mistake

in the

Wither

section. Finally, I
thank

Ulrich

me a Kantian; it taught me
a little harder to make myself

Mohrhoff for calling

that I
should work
look Jamesian.

References
[1] C. A. Fuchs, Quantum
of Quantum Information,
Implications

Information

in

Transfer:

Foundations

in the Light

inDecoherence and its

Quantum

and

Computation

Proceedings

of

the

NATO

Advanced

Research Workshop, Mykonos Greece, June 2530,

2000, edited by A.Gonis and P.E.A.Turchi (IOS

Press, Amsterdam, 2001),

pp. 3882.

Also posted

at quant-ph/0106166.
[2] C. A. Fuchs, Quantum
Are They?
posted

(The Post-V axj

at

States: What the Hell

o Phase

Transition),

http://netlib.bell-labs.com/who/cafuchs/.
[3] J. T. Cushing,

A. Fine, and

editors, Bohmian Mechanics

S. Goldstein,

and Quantum Theory:

An

Appraisal, (Kluwer, Dordrecht, 1996).

[4] R.B. Griffiths and R. Omn` es, Consistent
and Quantum Measurements,

Histories

Phys. Today 52(8),

2631 (1999).
[5]

J. G. Cramer,

Transactional
Mechanics,

An

of

Overview

Interpretation

of

the

Quantum

Int.J.

Theor. Phys. 27,227236 (1988).

[6] G. C. Ghirardi
Reduction

and

P. Pearle,

Theories: Changing

Dynamical

Quantum

so

Theory

the
Represents

Statevector

Proceedings

of

the

Reality,

Biennial

PSA

Meeting

1 990:
the

of

Philosophy

of Science Association, Vol. I,edited by A. Fine,

M.Forbes, and L.Wessels (Philosophy of Science

Association, East Lansing, MI, 1990),

pp. 1934.

53
[7] W. H. Zurek, Decoherence,

Einselection

and

the Existential Interpretation (The Rough Guide),

Phil. Trans. R. Soc. Lond. A 356, 17931821

(1998).

[8] P. Grangier, Contextual

Objectivity: A Realistic

Interpretation of Quantum Mechanics, quant-ph/

0012122;
Formalism

P. Grangier,

of

Reconstructing
Mechanics

Quantum

in

the
the

Contextual

Point of View, quant-ph/0111154.

Objectivity

[9] D.Deutsch, The Fabric of Reality: The Science

of Parallel Universesand

its Implications, (Allen

Lane, New York, 1997).

[10] L.Vaidman, The Many-Worlds
Quantum Mechanics,

of Philosophy

Interpretation

(Summer

at

http

://plato

edu/entries/qm-manyworlds/.
[11] C. Rovelli,

2002 Edition), edited by

E.N.Zalta (Stanford University, Stanford, CA, 2002).

Available

of

inThe Stanford Encyclopedia

Relational

Quantum

stanford

Mechanics,

Int. J.Theor. Phys. 35, 16371678 (1996).

[12] R. W. Clark, Einstein:

The Life and Times,

(World Publishing Co., New York, 1971). See Chapter

10:

...

[T]hey created the Study Group of German Natural

Philosophers

which

had at its disposal

large

sums of money, offered fees to those who would write

or speak against Ein-stein, and advertised its meetings

by large posters

concerts.

public

Company,

as

...
as

though there

In August

were

announcing

the Antirelativity

twenty

Einstein called it, announced

meetings to be held in Germanys biggest towns.

[13]

M.

Relativity

Gardner,

the

for

Million,

(Macmillan, New York, 1962).

[14] B. S. DeWitt
Letter

IQM-1

on

and R. N. Graham,
the

Interpretation

of

Resource
Quantum

Mechanics, Am. J.Phys. 39,724738 (1971).

[15] W. James, The Sentiment

in

of Rationality,

The Writings of William James: A Comprehensive

Edition, edited with

an

introduction

by J. J.

McDermott (Modern Library, Random House, New

York, 1967),

pp. 317345.

[16] C. W. Misner, K. S. Thorne, and J. A. Wheeler,

Gravitation, (W. H.Freeman, San Francisco, 1973),

p.23.
[17]

A. Zeilinger,

Quantum

Mechanics,

Foundational
Found.

Principle

Phys.

for

29, 631643

(1999);

C. Brukner
Invariant

and

Information

A. Zeilinger,
in Quantum

Operationally
Measurements,

Phys.
Rev. Lett. 83, 33543357

(1999); C. Brukner and

A.Zeilinger, Conceptual Inadequacy of the Shannon

Information

in Quantum

Phys.

Measurements,

Rev. A 63,022113-1022113-10

(2001); C.Brukner,

M. Zukowski, and A. Zeilinger, The Essence of

Entanglement,

Comment
Shannon

...

quant-ph/0106119; M. J.W. Hall,

on

Information

Zeilinger,

Conceptual

quant-ph/0007116;

Supposed

C. G. Timpson,

Inadequacy

Conceptual

Inadequacy

by C. Brukner

of

On

the

of

and A.

the

Shannon

Information, quant-ph/0112178.

[18] I.Devetak and A. E. Staples, Towards

Unification of Physics and Information Theory,

quant-ph/0112166.
[19] N. D.Mermin, The Contemplation
Computation,

of Quantum

Phys. Tod. 53(7), 1112 (2000).

[20] H. Barnum, Quantum

Information

Processing

and Quantum Logic: Towards Mutual Illumination,

quant-ph/0205129.
[21] P. Grangier, private communication, 2001.

[22] C. H. Bennett, T.A.Brun, C. M. Caves, and N.

D.Mermin, various vibes, 2001.

[23] C. H. Bennett

G. Brassard,

C. Cr epeau, R.

Jozsa, A. Peres, and W. K. Wootters, Teleporting

an
Unknown Quantum State via Dual Classical and

Einstein-Podolsky-Rosen Channels, Phys. Rev. Lett.

70, 1895 (1993).

[24] J. Bub, Quantum

Theory,

Mechanics

as a

Principle

Stud. Hist. Phil. Mod. Phys. 31, 7495

(2000). [25] J.Bub, Interpreting

the Quantum World,

(Cambridge University Press, Cambridge, 1997).

[26]

J. S.

Bell,

Speakable

Quantum Mechanics:

Collected

and

Unspeakable

Papers

on

in

Quantum

Philosophy (Cambridge U.Press, Cambridge, 1987).

[27] M. A. Nielsen and I. L. Chuang,

Computation

and

Quantum

Quantum

Information,

(Cambridge
University Press, Cambridge, 2000).

[28]

P.

Busch,

Valuations:

Giving

Quantum

Operational

States

as

Content

Effect

to

von

Neumanns
No-Hidden-Variables

Theorem,

quant-ph/9909073.

[29] C. M. Caves, C. A. Fuchs, K. Manne, and J.

Renes,

Gleasons

Theorem

with

POVMs,

unpublished,

2000.
[30]

A. M. Gleason,

Subspaces of

a Hilbert

on

Measures

the

Closed

Space, J.Math. Mech. 6,885

894
(1957).

[31] K. Vogel and H. Risken, Determination

Distributions

Quasiprobability

in

of

Terms

of

Probability
Distributions
Phys.

Phase,

the

for

Rev.

Rotated

40,

2847

Quadrature

(1989);

U.

Leonhardt,
Quantum-State

Tomography

and

Discrete

Wigner Function, Phys. Rev. Lett. 74,4101 (1995).

[32] C. M. Caves, C. A. Fuchs and R. Schack,

Unknown

Quantum

States:

The

de

Quantum

Finetti
Representation,

to

appear

in J. Math. Phys.

(2002). See quant-ph/0104088.

[33] R. Schack,
Quantum

T. A. Brun,

Bayes Rule,

and

C. M. Caves,

Phys. Rev. A 64, 014305-1

014305-4
(2001).

[34] A. Einstein, B. Podolsky, and N. Rosen, Can

Quantum-Mechanical Description of Physical Reality
Be

Considered

Complete?,

Phys.

Rev.

47,

777780 (1935).
[35] A.Fine, The Shaky Game: Einstein Realism and
the Quantum Theory, (University of Chicago Press,
1986);

Chicago,

Problem

in

Symposium

Its

M.

Jammer,

Historical

EPR

The

Development,

in

on
of Modern Physics: 50

the Foundations

years

of

the Einstein-Podolsky-Rosen Gedankenexperiment,

edited by P. Lahti and P. Mittelstaedt
Scientific, Singapore, 1985),

(World

pp. 129149.

[36] L. P. Hughston, R.Jozsa, and W. K. Wootters,

A Complete Classification of Quantum Ensembles

Having

a Given

Density Matrix,

Phys. Lett. A

183, 1418 (1993).

[37]

J. Bernstein,

Quantum

Profiles,

(Princeton

University Press, Princeton, NJ, 1991).

[38] A. Einstein, Remarks

Concerning

the Essays

Brought Together inthis Co-Operative Volume, in

Albert

Einstein:

Philosopher-Scientist,

third

edition, edited by P. A. Schilpp (Open Court, La

Salle,

IL,1970),
[39]

pp. 665688.

A. Peres,

Quantum

Theory:

Concepts

and

Methods, (Kluwer, Dordrecht, 1993).

[40]

C. M. Caves

Information:

How

and
Much

C. A. Fuchs,
Information

Quantum

in

State

Vector?, in

The Dilemma of Einstein, Podolsky and Rosen

60 Years Later, edited by A.Mann and M. Revzen,

Ann. Israel Phys. Soc. 12, 226257 (1996).

[41] W. Pauli, letter to M. Fierz dated 10 August

1954, reprinted and translated inK.V.Laurikainen,

Beyond the Atom: The Philosophical

Thought

of Wolfgang Pauli, (Springer-Verlag, Berlin, 1988),

p.226.
[42]

R.

Penrose,

Concerning

The

Computers,

Emperors

New

Mind:

Minds, and the Laws of

Physics,
(Oxford

University

Press,

1989),

Oxford,

pp.

268269.

[43] L. J. Savage, The

Foundations

of Statistics,

(Dover, New York, 1954).

[44] M. Jammer, The Experiment

in Classical and

inQuantum Physics, inProceedings of the International

Mechanics

Symposium,

Foundations

of

Quantum

in the Light of New Technology,

edited

by
S. Kamefuchi,

Namiki,

H. Ezawa,

S. Nomura,

Y. Murayama,

Y. Ohnuki,

and

M.

T. Yajima

(Physical
Society of Japan, Tokyo, 1984),

[45]

W.

Heisenberg,

Physics

pp. 265276.
and

Beyond:

Encounters

and Conversations,

translated

by A. J.

Pomerans
(Harper & Row, New York, 1971),

pp. 6364.

[46] C. A. Fuchs and A. Peres, Quantum

Needs No Interpretation,

;Quantum

(2000)

Formulation,

Inspiration:

Theory

Phys. Today 53(3), 70

Theory
Fuchs

and

Interpretation,
Peres

Reply,

53(9),

14 (2000).
[47] P. Simon, Kathys

Song,

(Columbia

Records,

New York, 1965).

55

[48] H. Barnum, C. M. Caves, J. Finkelstein, C. A.

Fuchs, and R.Schack, Quantum Probability from

Decision Theory?, Proc. Roy. Soc. Lond. A 456,

11751182 (2000).
[49] R. Cooke, M. Keane, and W. Moran,

Elementary

Proof

of Gleasons

Phil.

Soc.

Theorem,

An

Math.

Proc.
Camb.

Pitowsky,

Infinite

98,

117128

and Finite Gleasons

(1981);

I.

Theorems

and the
Logic

of Indeterminacy,

218228 (1998).

J. Math. Phys.

39,

[50]

K.

Kraus,

States,

and

Effects,

Fundamental Notions of Quantum

Operations.

Theory

(Springer,

Berlin,
1983), Lecture Notes inPhysics Vol. 190.
[51] A. S. Holevo, Information-Theoretical

Aspects

Prob. Info. Trans. 9,

of Quantum Measurement,

110118 (1973).
[52] S. J.
Strategies,

van

Enk, Quantum

and Classical Game

Phys. Rev. Lett. 84,789 (2000).

[53] C. A. Fuchs and J. Renes, Kochen-Specker-like

Noncolorability for Trine Measurements

ona Qubit,

long overdue, but still inpreparation.

[54] D. A. Meyer,
Nullifies

Finite

the Kochen-Specker

Precision

Measurement
Phys. Rev.

Theorem,

Lett.

83,

37513754

Non-Contextual

(1999);

Hidden

A.

Variables

Kent,

and

Physical

Measurements,

Phys. Rev. Lett. 83 37553757 (1999).

[55] I
thank Howard Barnum for pointing out the

similarities between this line of thought and that in

G. W. Mackey, Mathematical

Foundations

of

Quantum Mechanics, (W. A.Benjamin, New York,

1963),
[56]

pp. 6263.

C. A. Fuchs,

Notes

on a

Paulian

Idea:

Foundational,

Historical,

Anecdotal

and

Forward-Looking
Thoughts

on the Quantum,

quant-ph/0105039.

[57] N. E.Baytch and C. A. Fuchs, Nonorthogonal

Variations of Gleasons Theorem inQuantum Me-

chanics, forthcoming.
[58]

L.

Hardy,

Theory

Quantum

From

Five

Reasonable Axioms, quant-ph/0101012.

[59] L.Hardy, Why Quantum Theory?, to appear

inProceedings of the NATO Advanced Research

Workshop

on

Modality,

Probability,

and Bell s

Theorem, edited by J.Butterfield and T.Placek (IOS

Press, Amsterdam, 2002), quant-ph/0111068.
[60] E.Prugove cki, Information-Theoretical
of Quantum Measurements,

Aspects

Int.J.Theor. Phys.

16,321(1977).
[61]

F.

E.

Observables,

Schroeck,

(1989); F.E.
Schroeck,

Jr.,

Int. J. Theor.

Jr.

Yields Informational

Coexistence

Phys.

Unsharpness
Completeness,

of

28, 247262

in Measurement
in Symposium

on
the
Quantum

Foundations
Theory

Philosophical

of Modern
of Measurement

Physics
and

1 990:
Related

Problems, edited by P. Lahti and P. Mittelstaedt

(World Scientific, Singapore, 1991),
Jr., B.

Schroeck,

Fischer,

and
and

Completeness

Informational

pp. 375389;

F.E.

G. Harnett,

Frames,

in

Symposium

on

the Foundations
Measurement,

Quantum

of

Modern

Physics

Irreversibility

1993:

and

the

Physics

of Information, edited by P.Busch, P. Lahti, and

P.Mittelstaedt (World Scientific, Singapore, 1993),

pp. 378389.
[62] P. Busch,

Physical

Complete

Informationally

Quantities,

Int.

J.

Theor.

of

Sets

Phys.

30,

12171227
(1991).

[63] C. A. Fuchs, L. Huelsbergen,

Investigations

on Informationally

and G. G. Plunk,

Complete Measure-

ments, inpreparation.
[64]

R.

Jozsa,

Computation,

Entanglement

in The Geometric

and

Quantum

Universe:

Science,

Geometry,

and the Work of Roger Penrose, edited by S. A.

Huggett, L.J.Mason, K.P.Tod, S.T.Tsou, and

N. M. J. Woodhouse, (Oxford U. Press, Oxford,

1998),

p.369.

[65] N. D. Mermin, Boojums All the Way Through:

Communicating Science in a Prosaic Age, (Cambridge

U.Press, Cambridge, 1990).

[66] R. Bhatia,

Matrix

Analysis,

(Springer-Verlag,

New York, 1997).

[67]

N. R. Wallach,

An

Unentangled

Gleasons

Theorem, quant-ph/0002058.

56
[68]

D.

Aerts

Justification

for

and

I. Daubechies,

Using

the

Tensor

Physical

to

Product

Describe Two
Quantum

Systems

as

One Joint System,

Helv.

Phys. Acta 51, 661675 (1978); D.Aerts, Subsystems

in Physics Described by Bi-linear Maps between
the Corresponding Vector Spaces, J.Math. Phys.

21, 778788 (1980); M. Kl

ay, C. Randall,

and D.

Foulis, Tensor Products and Probability Weights,

Int. J. Theor.
Kl

ay,

EPR

Phys. 26, 199219

Experiments:

The

(1987); M.

Structure

of

the

Sample

Space,

Found. Phys. Lett. 1, 205243

(1988);

A.Wilce, Tensor Products inGeneralized Measure

Theory,
(1992).

Int. J. Theor.

Phys. 31, 19151928

[69] C. H. Bennett, D.P. DiVincenzo, C. A.Fuchs, T.

Mor, E.Rains, P.W. Shor, J.A.Smolin, and W. K.
Wootters,
Entanglement,

[70]

S.

Nonlocality

Quantum

without

Phys. Rev. A 59,10701091 (1999).

MacLane

and

G.

Birkhoff,

Algebra,

(Macmillan, New York, 1967).

[71] H. Araki, On

Characterization

Space of Quantum Mechanics,

of the State

Comm. Math. Phys.

75,
124

(1980);

Accessibility

of

W.

K.

Quantum

Wootters,

Local

Information,

in

Complexity, Entropy

and the Physics of Information, edited by W. H.

Zurek (Addison-Wesley, Redwood City, CA, 1990),

p.2946.
[72] A. Wilce, Tensor

Product of Frame Manuals,

Int. J.Theor. Phys. 29, 805814 (1990)

[73] R. S. Cohen, M. Horne, and J. Stachel, editors,
Potentiality, Entanglement and Passion-at-a-Distance,

(Kluwer, Dordrecht, 1997).

[74] C. M. Caves,
Bayesian

C. A. Fuchs, and R. Schack,

Probability

in Quantum

Mechanics,

in

preparation (2002).
[75] H. E. Kyburg, Jr. and H. E. Smokler, eds., Studies

in Subjective Probability, Second Edition (Robert E.

Krieger Publishing, Huntington, NY, 1980).

[76] E. T. Jaynes,

Probability

Theory:

of Science, available for download

on

The Logic

the World Wide

Web

at http://bayes.wustl.edu/.

was unfortunately never

This huge book

finished due to Prof. Jaynes

as perhaps

death in 1998. Nevertheless, it stands

the most complete and persuasive document inthe

field.
[77] J. M. Bernardo and A. F. M. Smith, Bayesian
Theory (Wiley, Chichester, 1994).
[78] N.D.Mermin, From Classical State-Swapping

Quantum Teleportation,

to

Phys. Rev. A 65,012320-

1012320-6 (2002).
[79]

B.

Schumacher,

Entanglement

Sending

Quantum

Through Noisy Channels,

Phys. Rev.

A 54,2614

2628 (1996).
[80] T. M. Cover
of Information

and J. A. Thomas,

Elements

Theory, (John Wiley & Sons, New

York,
1991).
[81] R. B. Ash, Information
York, 1965).

Theory, (Dover, New

[82] A. Wehrl, General Properties of Entropy,

Rev.

Mod. Phys. 50,221259 (1978).

[83] W. K. Wootters,

Random

Quantum

States,

Found. Phys. 20,13651378 (1990); K.R.W. Jones,

Riemann-Liouville

Fractional

Integration

on Hyperspheres,

Reduced Distributions

and

J.Phys. A

24, 12371244 (1991).

[84] R. Jozsa, D. Robb, and W. K. Wootters,

Lower

Bound

Accessible

for

in

Information

Quantum
Mechanics,

[85]

Phys. Rev. A 49,668677 (1994).

C. A. Fuchs

Tradeoff

Relations

and
for

K. Jacobs,

Information

Finite-Strength

Quantum

Measurements, in Phys. Rev. A 63, 062305-1062305-15

(2001).
[86]

M. A. Nielsen,

Measurement

Mixing

Characterizing

in Quantum Mechanics,

and

Phys. Rev. A

63,
022114-1022114-11
[87]

R.

Schatten,

(2001).

Norm

Ideals

of

Completely

Continuous Operators, (Springer-Verlag, Berlin, 1960)

[88] H. M. Wiseman, Using Feedback

Back-Action in Quantum Measurements,

to Eliminate
Phys. Rev.

51, 24592468

(1995).;

J. Audretsch,

Di osi, and T.Konrad, Evolution

of

a Qubit

L.

under

the

Influence

of

Measurements,

Succession

Unsharp

of

quant-ph/0201078.

57
[89] J. Bub, Von Neumanns

as a Probability

Projection

Postulate

Conditionalization Rule in Quantum

Mechanics, J.Phil. Logic 6,381390 (1977)

[90] S. L. Braunstein and C. M. Caves, Wringing

Out Better Bell Inequalities,

Ann. Phys. 202, 2256

(1990).

[91] M. Ozawa, Quantum

State Reduction and the

Quantum Bayes Principle, inQuantum Communication, Computing,

and Measurement,

by O. Hirota, A. S. Holevo,

and

edited

C. M. Caves

(Plenum

Press,
Ozawa,

An

New

York,

Operational

1997),

pp.

Approach

233241;

M.

to Quantum

State Re-

duction, Ann. Phys. 259, 121137 (1997); M.

Ozawa, Quantum State Reduction: An Operational

Approach, Fortschr. Phys. 46, 615625 (1998).

[92] R. Duvenhage,

The Nature of Information

Quantum Mechanics, quant-ph/0203070.

in

[93] B. C. van Fraassen, private communication,

11

April 2002.
[94] G. Boole, An Investigation

of the Laws

of

Thought, (Dover, New York, 1958).

[95] T.A. Brun, J.Finkelstein, and N.D.Mermin,

How Much State Assignments Can Differ, Phys.

Rev. A 65,032315-1032315-3

(2002).

[96] C. M. Caves, C. A. Fuchs, and R. Schack,

Compatible

Quantum States: Coming to Agreement

and
Uncovering Disagreement,

inpreparation.

[97] C. M. Caves, private communication,

as 10 September

as

early

1999.

[98] G. Vidal and J.I.Cirac, Storage of Quantum

Dynamics inQuantum States: A Quasi-Perfect Programmable

quant-ph/0012067;

Quantum

W. D

ur

Gate,

and

J. I. Cirac,

Non-local Operations:

Purification,
Tomography,

and

Storage,
Probabilistic

Compression,
Implementation,

Phys. Rev. A

64, 012317-1012317-14

(2001); S. F. Huelga, J.

A.Vaccaro, A.Chefles, and M. B.Plenio, Quantum

Remote
Operations,

Vi-

Control:

Teleportation

of

Unitary

Phys. Rev. A 63, 042303-1042303-4;

G.

dal, L. Masanes,

and

J. I. Cirac,

Storing

Quantum Dynamics in Quantum States: A Stochastic

Pro-

grammable Gate, Phys. Rev. Lett. 88, 047905-1

047905-4 (2002); C.-P. Yang and J. Gea-Banacloche,

Teleportation
Receiver-Encoded

Rotations

of
Secret

Sharing,

and
J. Opt.

B 3,

407411 (2001); S. F.
Huelga,

Remote

M. B. Plenio,

Control

and

of Restricted

J. A. Vaccaro,

Sets of Operations

Teleportation

of

[99]

Phys.

Angles,

042316-1042316-12

Rev.

65,

(2002).

A. Jamio lkowski,

Linear

Transformations

which Preserve Trace and Positive Semidefiniteness

of
Operators,

Rep.

Math.

Phys.

3, 275278

(1972).

[100] M.-D. Choi, Completely

on Complex

Matrices,

Positive Linear Maps

Lin. Alg. App. 10,285290

(1975).

[101] P. W. Shor, Scheme for Reducing Decoherence

inQuantum Computer Memory, Phys. Rev. A 52,

R2493 (1995); A. M. Steane, Error Correcting
Codes in Quantum Theory, Phys. Rev. Lett. 77, 793

(1996).

[102] C. H. Bennett
Cryptography:

and G. Brassard,

Public

Key

Distribution

Quantum

and

Coin

Tossing,

in Proc.

Computers,

IEEE

International

on

Conference

and Signal Processing

Systems

(IEEE

Press,
New York, 1984),

p. 175, IEEE,

Bennett, Quantum Cryptography

Nonorthogonal

States,

1984; C. H.

Using Any Two

Phys. Rev. Lett. 68,

3121 (1992).

van Enk and C. A. Fuchs, Quantum

an Ideal Propagating Laser Field, Phys.

[103] S. J.
State of

Rev.
Lett. 88, 027902-1027902-4
Enk

and

C. A. Fuchs,

(2002); S. J.

Quantum

of

State

van
a

Propagating
Laser Field,

Quant. Info. Comp. 2, 151165

(2002).
[104] B. de Finetti, Theory of Probability
New York, 1990); For

a collection

(Wiley,

de Finettis origi-

nal papers and their translations into English,

see P.Monari and D. Cocchi, editors, Probabilit` a e
InduzioneInduction

and

Probability

(Biblioteca di Statistica, CLUEB, Bologna, 1993).

[105] R. L. Hudson
Normal

Symmetric

and G. R. Moody,

States

and

an

Locally

Analogue

of de

Finettis

verw.

Z. Wahrschein.

Theorem,

Geb.

33,

343351 (1976).
[106] E. C. G. Stueckelberg,

Real

Hilbert

Quantum

Helv. Phys.

Space,

Theory

Acta

in

33, 727

(1960);
S. L.Adler, Quaternionic

Quantum Mechanics

and Quantum Fields (Oxford U.Press, New York,

1995).

58
[107] C. H.Bennett, G. Brassard, and N.D.Mermin,
Quantum Cryptography

without Bells Theorem,

Phys. Rev. Lett. 68, 557559

Fuchs, Information

Gain

vs.

(1992); C. A.

State Disturbance

in

Quan-

tum Theory,

Fort.

der Phys.

46, 535565

(1998).
[108] C. A. Fuchs

vs.

Disturbance

and A. Peres, Quantum

Information

Gain:

State

Uncertainty

Relations

for Quantum

20382045

(1996);

Information,

C. A. Fuchs,

Phys. Rev. A 53,

N. Gisin,

R. B.

Griffiths,

C.-S.

Niu,

Eavesdropping

in

and

A.

Quantum

Peres,

Optimal

Cryptography.

I.

Information Bound
and

Optimal

Strategy,

(1997);

11631172

D.

Phys.

Bruss,

Rev. A 56,

Optimal

Eavesdropping in

Quantum

with

Cryptography

Phys. Rev. Lett. 81, 25982601

Six

States,

(1998); K. Banaszek

and
I. Devetak,
Ensembles

Fidelity

of Identically

for

Finite

Qubits,

Phys.

Trade-off

Prepared

Rev. A

64, 052307-1052307-14 (2001).

[109] R. W. Spekkens and T.Rudolph, Degrees

of

Concealment and Bindingness inQuantum Bit Com-

Phys. Rev. A 65, 012310-1

mitment Protocols,

012310-10 (2001); R.W. Spekkens and T.Rudolph,

of

Optimization
Generalizations
Commitment

Coherent

the

of

BB84

Attacks

Quantum

in
Bit

Proto-

col, Quant. Inf. Comp. 2, 6696 (2002); R. W.

Spekkens and T.Rudolph, A Quantum Protocol for

Weak

Cheat-Sensitive

quant-ph/0202118;

Coin

Flipping,

L. Hardy and A. Kent, Cheat

Sensitive
Quantum

Bit

Commitment,

quant-ph/9911043.
[110] R. Blume-Kohout,

C. M. Caves, and I. H.

Deutsch,

Mount

Climbing

Scalable:

Physical-Resource
Requirements

[111]

for

Scalable

Quantum

quant-ph/0204157.

Computer,

E. T. Jaynes,

Probability,

The

in Appendix

de Finetti

System

of

A, Other Approaches

to

Probability Theory, of Ref. [76].

[112]

M. Fortun

Through:

and

Pursuing

H. J. Bernstein,

Science

and

Muddling

Truths

in

the

Twenty-First
Century,

1998).

59

(Counterpoint,

Washington,

DC,