lmps plenary 2011

OverviewBasic framework

Main exampleDiscussion

Retrocausality – What Would it Take?

Huw Price

Title First LastnamePositionFaculty

The University AlumniExcellence Award 2011

This award is presented to

Firstname LastnameIn recognition of

outstanding achievement

Huw Price Retrocausality – What Would it Take? 1/30

1 Brief overview

2 The basic framework

3 The main example

4 Some discussion

The issues

1 Could we have good scientific evidence for thinking that we can influencethe past?

2 If so, what would such evidence look like?

3 Do we have any such evidence in fact?

The issues

Structure of the talk

1 Simplifying assumptions – making the issue as simple as possible, bysetting aside various metaphysical complexities.

(Would the complexities make a difference? Maybe, maybe not, but let’sconsider the simple case first!)

2 The central example – showing how some of the distinctive correlationsof QM can commit us to retrocausality, in the defined sense.

(The argument depends on three further assumptions, which – though notincontestable – are far from implausible. Thus it shows how, in accepting thoseassumptions, we might take ourselves to have discovered retrocausality.)

3 Discussion.

The block universeKnowledge and choiceWhy not retrocausality?Interventionism

1 Brief overview

2 The basic frameworkThe block universeKnowledge and choiceWhy not retrocausality?Interventionism

3 The main example

4 Some discussion

The block universe

[Images: Scientific American/Bryan Christie Design]

The block universe

Choice & ignorance: the agent’s perspective

As I decide how to act, I’m ignorantabout what I will do. Will I do A or -A?When I decide I’ll know, but not before!

My ignorance extends to whatever I taketo depend on my choice . . . normally inthe future.

Given the block universe, this is ignoranceabout how the future actually is. It relieson hypotheticals, not counterfactuals:

If I (actually) do A, then E(A)If I (actually) do -A, then E(-A).

E(A) E(-A)

Choice-dependentignorance

Outcome

E(A) E(-A)

Outcome

E(A) E(-A)

Outcome

E(A) E(-A)

Outcome

E(A) E(-A)

Outcome

Why not backwards?

Why can’t my zone of choice-dependentignorance extend into the past, as well asthe future . . . so that my decisionhypotheticals look like this?

It can’t include the ‘known’ (or knowable)past, of course, because that conflictswith ignorance (I could ’bilk’) – but whyshould the entire past be assumed‘knowable’, for these purposes?

This is what I mean by ‘retrocausality’.

Our question: What would it take toshow that our choice-dependentignorance extends into the past?

E(A) E(-A)

Outcome

Why not backwards?

E(A) E(-A)

Outcome

Why not backwards?

E(A) E(-A)

Outcome

Why not backwards?

Outcome

Why not backwards?

Outcome

Why not backwards?

E(A) E(-A)

Outcome

Why not backwards?

E(A) E(-A)

Outcome

Why not backwards?

E(A) E(-A)

Outcome

Interventionism

Assuming determinism, ignorance aboutmy choice is bound to imply ignoranceabout something in my own past.

I’ll ignore this, and assume an‘interventionist’ model of agency – i.e,that our actions come from ‘outside’ thesystems we consider. (Cf. Pearl)

So the only issue is this one: what wouldit take to convince us that this is actuallythe right model of our situation?

E(A) E(-A)

Outcome

Interventionism

E(A) E(-A)

Outcome

Interventionism

E(A) E(-A)

Outcome

Interventionism

E(A) E(-A)

Outcome

Interventionism

E(A) E(-A)

Outcome

Interventionism

E(A) E(-A)

Outcome

Classical Polarization 101Connection and control AQuantum Polarization 101Connection and control B

1 Brief overview

3 The main exampleClassical Polarization 101Connection and control AQuantum Polarization 101Connection and control B

4 Some discussion

The classical case (right)

R = 1τ

Polarizing cubeset at angle σR

L = 1τL

Photon

Polarizing cubeset at angle σL

Malus’ Law: IntensityR=1 = cos2(τ − σR ), IntensityR=0 = sin2(τ − σR )

The classical case (right)

R = 1τ

L = 1τL

Photon

Malus’ Law: IntensityR=1 = cos2(τ − σR ), IntensityR=0 = sin2(τ − σR )

The classical case (left)

R = 1τR

Photon

L = 1τ

‘T-reversed’: IntensityL=1 = cos2(τ − σL), IntensityL=0 = sin2(τ − σL)

The classical case (left)

R = 1τR

Photon

L = 1τ

‘T-reversed’: IntensityL=1 = cos2(τ − σL), IntensityL=0 = sin2(τ − σL)

Thinking about connection and control

L = 1τ

1 Polarization as a ‘local’ mechanism

2 What we can wiggle

L = 1τ

Defeating the Demon of the Left?

L = 1τ

Suppose we control the left polarizer angle, σL, but not the input beams –the Demon of the Left controls those. Can we control the polarization of theoutput beam, τ?

No! The Demon can make τ whatever He likes, by choosing the appropriateinput intensities. (The Demon can ‘counteract’ our wiggles, in other words.)

L = 1τ

Nature as the Demon of the Right

L = 1τ

Similarly, suppose we control the right polarizer angle, σR , but not theoutput beams – Nature controls those. Can we control the polarization of theinput beam, τ?

No. Nature absorbs our wiggles in changes in the output intensities, and τdoesn’t change. (So no retrocausality here!)

L = 1τ

The quantum case (right)

R = 1τ

Photon

L = 1τL

Photon

QM Malus’ Law: ProbR=1 = cos2(τ − σR ), ProbR=0 = sin2(τ − σR )

The quantum case (right)

R = 1τ

Photon

L = 1τL

Photon

QM Malus’ Law: ProbR=1 = cos2(τ − σR ), ProbR=0 = sin2(τ − σR )

The quantum case (left)

R = 1τR

Photon

L = 1τ

Photon

‘T-reversed’: ProbL=1 = cos2(τ − σL), ProbL=0 = sin2(τ − σL)

The quantum case (left)

R = 1τR

Photon

L = 1τ

Photon

‘T-reversed’: ProbL=1 = cos2(τ − σL), ProbL=0 = sin2(τ − σL)

Connection and control again

L = 1τL

Photon

1 Polarization as a local mechanism – intuitively, τL is just whatever‘beable’ connects changes on the left to changes on the right. (LetRealism be the assumption that here is some such beable.)

2 What we can wiggle?

L = 1τL

Photon

L = 1τL

Photon

L = 1τL

Photon

L = 1τL

Photon

L = 1τL

Photon

Defeating the Demon of the Left

L = 1τL

Photon

Again, suppose we control the left polarizer, σL, but not the input photons– the Demon of the Left controls those. Can we control τL?

Yes! . . . so long as the Demon has to put the photon on one channel orother (the Discreteness assumption). In this case, τL = σL or τL = σL + π/2.So we can’t fix τL completely, but we can restrict it to just two possibilities.

L = 1τL

Photon

L = 1τL

Photon

L = 1τL

Photon

L = 1τL

Photon

L = 1τL

Photon

L = 1τL

Photon

Defeating Nature on the right

L = 1τL, τR

Photon

Suppose we control the right polarizer, σR , but not the output beams –Nature controls those. Can we control some property τR of the input beam?

Yes, on two assumptions: (1) Time-symmetry: if τL exists, so does τR

(with T-reversed rules); (2) Discreteness: Nature has to put the output photonon one channel or other. Then τR = σR or τR = σR + π/2 – we can’t fix τR

completely, but we can restrict it to two possibilities. (Retrocausality!)

L = 1τL, τR

Photon

L = 1τL, τR

Photon

L = 1τL, τR

Photon

L = 1τL, τR

Photon

L = 1τL, τR

Photon

L = 1τL, τR

Photon

L = 1τL, τR

Photon

L = 1τL, τR

Photon

L = 1τL, τR

Photon

More about DiscretenessRemembering the assumptionsRewriting Bell?Postscript

1 Brief overview

3 The main example

4 Some discussionMore about DiscretenessRemembering the assumptionsRewriting Bell?Postscript

The status of the Discreteness assumption

L = 1τL

Photon

Discreteness is hardly a realistic assumption on the left: plenty ofunDemonic real-world experimenters know how to supply photons to thisexperiment in a superposition of L = 0 and L = 1.

L = 1τL

Photon

Discreteness is hardly a realistic assumption on the left: plenty ofunDemonic real-world experimenters know how to supply photons to thisexperiment in a superposition of L = 0 and L = 1.

L = 1τL

Photon

But on the right we can force Nature to be discrete, simply by making ameasurement . . . unless there is some non-discrete ontology (e.g., anuncollapsed ontic wave function) that survives measurement.

So realist ‘no-collapse’ theories (e.g., dBB and Everett) automatically evadethe argument. Otherwise, measurement seems to guarantee Discreteness.

L = 1τL

Photon

L = 1τL

Photon

L = 1τL

Photon

Framework assumptions

Three simplifying assumptions

1 The block universe

2 ‘Hypotheticals not counterfactuals’

3 An interventionist model of ‘causal influence’

Would relaxing these assumptions make a difference?

I’m skeptical, but I’m not making any claims about this issue here.

Foreground assumptions

L = 1τL, τR

Photon

1 Realism: We assumed that τL is a beable (providing a mechanism toexplain how changes on the left ‘produce’ changes on the right.)

2 Time-symmetry: This is what gets us from the existence of τL to theexistence of τR .

3 Discreteness: Apparently guaranteed by measurement, as above, except in‘no collapse’ theories. (Note we don’t need it ‘every time’, but only ‘on demand’,for the argument to go through.)

L = 1τL, τR

Photon

L = 1τL, τR

Photon

L = 1τL, τR

Photon

L = 1τL, τR

Photon

What we get from this case

We have offered an argument for this:

(QM) ` (Realism ∧ T-Symmetry ∧Discreteness)⇒ Retrocausality

By transposition, we get:

(QM) ` ¬Retrocausality⇒ ¬(Realism ∧ T-Symmetry ∧Discreteness)

So if you had a good argument against retrocausality, you could derive thefollowing nice result:

(QM) ` ¬(Realism ∧ T-Symmetry ∧Discreteness)

[Theorem]

Even better, why not just follow Bell’s Theorem and various No HiddenVariable theorems, and simply assume that there is no retrocausality? Then wealready have a theorem!

My recommendation: Play the opposite hand – take the initial result as amotivation for questioning the standard assumption of Bell’s Theorem, etc.

[Theorem]

(QM) ` ¬(Realism ∧ T-Symmetry ∧Discreteness) [Theorem]

[Theorem!]

Rewriting Bell’s Theorem?

The usual formulation

Locality + ‘Independence’ (= ¬Retrocausality) ` ¬QM

Equivalently

QM + ¬Retrocausality ` Nonlocality

Transposing

QM + Locality ` Retrocausality

. . . so here’s another answer to our original question!

Equivalently

Transposing

Equivalently

Transposing

Equivalently

Transposing

Equivalently

Transposing

Retrocausality – what would it take?

Our answer

A world in which the laws somehow make it impossible for the variationwe have at our disposal as agents to be fully absorbed in future, so thatsome of it has to be absorbed in the past.

For example, the world we find in QM, so long as it is interpretedrealistically, and assumed to be time-symmetric; and so long as thediscreteness characteristic of quantum phenonema is assumed to befundamental.

The answer implicit in Bell’s Theorem

The world we find in QM, so long as it is assumed to be local.

Our answer

Though not to Bell’s taste!

Motivations and history

44 O. C O S T A D E B E A U R E G A R D

This is t h e i n t e r p r e t a t i o n of t h e E i n s t e i n p a r a d o x I (12) have p r o p o s e d qu i t e a few t imes , a n d which now STAPP (~3), BELL (~4), I)AVIDON (la) ,~nd o the r s (~G) are more or less a d v o c a t i n g or p o i n t i n g at .

Fig. 2. - Space-time diagram of the Einstein paradox: the die is cast not at the severance point- ins tant C, but later, at L and =V, where and when the measurements are performed. Thus the correlation between L and .Y is t ied through C, in their common past , cia the Feynman-s ty le zigzag made of the t imelike vectors CL and CN.

2. - Ncoquantal and paleoquantal calculat ions for cascades.

F r o m t h e two (o r thogona l ) p u r e he l i e i ty s t a t e s L,,Lb a n d R~R~ of t h e pa i r s of p h o t o n s a a n d b, we bu ih t t h e two (o r thogona l ) P - i n v a r i a n t s t a t e s

J[ 1 ~z -z (L, Lb + R,R~) = .~ (E~,E~ q- T~,Eo) ,

(L~Lb-- R,,Rb) ., [L , ,Eo-- E,~E;] ,

where t h e we l l - kn own f o r m u l a e

(4) ~+/;2 L~ = E ; - - iE; ,

+/'2 Rb = E; -+- iE;

h a v e been u s e d ; y a n d z d e n o t e a r b i t r a r y Car t e s i an axes o r t h o g o n a l to t h e l ine of f l ight x of t h e two p h o t o n s .

(12) O. COSTA DE BEAUREGAI/D: Compt. Rend., 236, 1632 (1953); Rev. Interu. Philos., 61-62, 1 (1962); Dialectica, 19, 280 (1965); in Proceedings o] the International Conjerence on Thermodynamics, edi ted by P. T. LANDSBERG (London, 1970), p. 539. (la) tI . P. STAPI': Nuaro Cimento, 29B, 270 (1975). (14) j . S. BELL: ~pist. Lett., 9, 11 (1976). Q~) W. C. DAVlDON: ~UOVO Cimento, 36 B, 34 (1976). (16) See footnote (13) in J. F. CLAVS]~R and M. A. HORNE: Phys. I~ec. D, 10, 526 (1974).

“Space-time diagram of the Einstein paradox . . .the correlation between L and N is tied throughC, in their common past, via the Feynman-stylezigzag made of the timelike vectors CL and CN.” O. Costa de Beauregard (1911–2007)

I have been using QM to provide an example, in thinking aboutretrocausality in general.

But one motivation for thinking about retrocausality in general, is that itmight have application to some of the puzzles of QM – e.g., that it mighttake the sting out of ‘nonlocality.’

Remembering the pioneer of this idea . . .

J[ 1 ~z -z (L, Lb + R,R~) = .~ (E~,E~ q- T~,Eo) ,

(L~Lb-- R,,Rb) ., [L , ,Eo-- E,~E;] ,

(4) ~+/;2 L~ = E ; - - iE; ,

+/'2 Rb = E; -+- iE;

J[ 1 ~z -z (L, Lb + R,R~) = .~ (E~,E~ q- T~,Eo) ,

(L~Lb-- R,,Rb) ., [L , ,Eo-- E,~E;] ,

(4) ~+/;2 L~ = E ; - - iE; ,

+/'2 Rb = E; -+- iE;

J[ 1 ~z -z (L, Lb + R,R~) = .~ (E~,E~ q- T~,Eo) ,

(L~Lb-- R,,Rb) ., [L , ,Eo-- E,~E;] ,

(4) ~+/;2 L~ = E ; - - iE; ,

+/'2 Rb = E; -+- iE;

J[ 1 ~z -z (L, Lb + R,R~) = .~ (E~,E~ q- T~,Eo) ,

(L~Lb-- R,,Rb) ., [L , ,Eo-- E,~E;] ,

(4) ~+/;2 L~ = E ; - - iE; ,

+/'2 Rb = E; -+- iE;

“Space-time diagram of the Einstein paradox . . .the correlation between L and N is tied throughC, in their common past, via the Feynman-stylezigzag made of the timelike vectors CL and CN.”

O. Costa de Beauregard (1911–2007)

J[ 1 ~z -z (L, Lb + R,R~) = .~ (E~,E~ q- T~,Eo) ,

(L~Lb-- R,,Rb) ., [L , ,Eo-- E,~E;] ,

where t h e w e l l - kn o wn f o r m u l a e

(4) ~+/;2 L~ = E ; - - iE; ,

+/'2 Rb = E; -+- iE;

The End

lmps plenary 2011

Education

discovered retrocausality

simple case

talk1 simplifying assumptions

good scientic evidence

central example

distinctive correlationsof

dened sense

inuencethe past