Delayed Choice exp and Cause-Effect

[Gerinski]

Main Question or Discussion Point

Amateur level.

Does the delayed-choice double slit experiment violate or put in trouble the standard cause-and-effect thinking?

 

[vanesch]

Does the delayed-choice double slit experiment violate or put in trouble the standard cause-and-effect thinking?

Yes.
Does it mean that there is no cause-effect relation anymore ? No.
What does it mean ? That however this riddle is solved, it is going to be subtle :-)

 

[Gerinski]

Maybe an extension of the question: "what does the delayed-choice experiment tells us about time?"
(I know there's no answer, just opinions :-)

 

[Sherlock]

Does the delayed-choice double slit experiment violate or put in trouble the standard cause-and-effect thinking?

No. Whichever detection method is implemented (at a
certain time corresponding to the quantum disturbance
being between the two-slit and the detector), it is in
place before the disturbance reaches it.

If single detectors, one focused on one slit and the
other focused on the other slit, are chosen, then
one or the other detector will register -- as if the
disturbance had gone through only one slit.

If a screen is chosen, then a single dot will appear
in one of the maxima regions of an (eventually apparent
if enough dots are accumulated) interference pattern --
as if the disturbance had gone through both slits.

Note the use of the term, "as if". In the case of the
single focused detectors, it just isn't known whether
the emitted disturbance went through both slits or
only one. If modern physics has taught us anything,
it's that just because you can't see something, doesn't
mean that there's nothing there.

In the case of the screen detector (which will
eventually produce an interference pattern built
up dot by dot), it also isn't known whether the emitted
disturbance went through both slits or only one.

Anyway, the experimental sequence of events doesn't
contradict standard causality. Neither does the data.
It's just so far resistant to a qualitative understanding
via classical imagery.

 

[mccrone]

Maybe an extension of the question: "what does the delayed-choice experiment tells us about time?"
(I know there's no answer, just opinions :-)

Remember that there are always two issues in epistemology - what is a useful model of something and what is the truth? A mechanical conception of causality and time can still be useful even when patently "untrue" - but yes QM nonlocality is proof that simple cause and effect thinking (which embodies the mechanical principle of locality) cannot be the whole story.

One approach to rethinking time would be Cramer's transactional interpretation - where particles make an offer wave that travels forward in time and some absorber particles accept the offer with a wave that travels back in time.

As a "model" the picture of a negotation going forwards and backwards in a single time dimension fits the nonlocal facts. However it is still tied to a mechanical logic and so not quite revolutionary enough for my tastes to be more than a stepping stone.

Another related way of making sense of things would be to take Feynman's sum over histories approach very literally. All possible paths (in space and time) between two particles exist. So there is a "thick" spatiotemporal description of the world that is the wavefunction and it exists all at once. Then there is the collapse to a "thin" actual path that creates a strand of linear, cause and effect, spacetime.

In this view, time would not be a single thin dimension but a hierarchical stack of temporal scales. The kind of time you see then depends on where you are located. If you get down to the most local scale, it will seem that you are stuck in a very small "now" - an instant - with past and future stretching out behind and before you. If you could see the world on a global scale, then the "now" would include that past and future. Like time for a photon, the world would seem frozen still with both ends of a trajectory already having happened.

Time is about change and here we are describing a spatiotemporal hierarchy of scales in which the smallest scale is a blur of events and the largest scales appear frozen still. In our Universe, the smallest "now" would be Planck scale, the largest would be (perhaps) the lightcone of the visible universe.

 

[Sherlock]

... QM nonlocality is proof that simple cause and effect thinking (which embodies the mechanical principle of locality) cannot be the whole story.

Quantum nonlocality has to do with global contexts, systems of correlated
spatially separated measurments rather than individual (local) measurements.
It hasn't altered our standard cause and effect thinking, or told us anything
more about time than was already known -- and neither has the delayed
choice double-slit experiment (re the original poster's question).

The time of an event is the configuration (the reading) of a clock
correlated with the event (which is itself a configuration of some
set of objective phenomena -- eg., a ball contacting a bat, a spot
of light appearing or disappearing, a fire igniting, a picture hanging
on a wall, a moving car passing a stop sign, a car sitting in a
driveway, etc.).

Time is the indexing of changing configurations. In an expanding
universe, each successive universal configuration is different than
it's immediate predecessor, but also more like it than any other
prior configurations. This is the way things seem to work from
the largest to the smallest scales. That is, there is a definite
direction to time (change). So, eggs don't spontaneously unfry,
broken cups don't spontaneously reassemble, waves move
away from (rather than toward) the disturbances that created
them, and so on. Time doesn't reverse because it can't
reverse in an expanding universe. Effects can't come before
causes because that would simply be misusing the terms.
The universe will never revisit it's past, and neither will
anything in it -- except in our imaginations and memories.

 

[mccrone]

Quantum nonlocality has to do with global contexts, systems of correlated
spatially separated measurments rather than individual (local) measurements.

It is to do with temporal as well as spatial correlations. So delayed choice twin slit experiments are a stark demonstration.

The time of an event is the configuration (the reading) of a clock
correlated with the event (which is itself a configuration of some
set of objective phenomena -- eg., a ball contacting a bat, a spot
of light appearing or disappearing, a fire igniting, a picture hanging
on a wall, a moving car passing a stop sign, a car sitting in a
driveway, etc.).

This might be how a human observer chooses to index events - motivated by a belief in cause and effect or locality. But nonlocality raises the question of when an event really occurs. The point I was making was that a global scale observer would see both ends of an event as part of the same effective moment. So both the photon emission by a star in some distant galaxy and its "much later" absorbtion by my eye.

What the proof of nonlocality tells us is that locality-based models of causality are incomplete (though they are certainly still useful). So I was talking about the kinds of models based on hierarchy theory that might present a different view of time.

Effects can't come before causes because that would simply be misusing the terms.

Yeah but then you have the problem of the chicken and the egg. Cause and effect thinking runs into paradoxes - like how can a something (the universe) spring out of a nothing.

So to get out of this, you have to look into other causal models. So for example, ones that start with a state of vague everythingness (cf: Anaximander, Peirce) and then dichotomise or symmetry-break to produce two crisp limits on being. So we would now start with a vague chicky-egginess and watch it divide asymmetrically into a chicken and egg (or if you like, the first egg inside the first chicken). So now causes are effects.

I realise that alternatives to locality and mechanistic logic are not fashionable. But non-locality has to be accounted for within some causal model unless you want this aspect of reality to remain a mystery.

 

[Sherlock]

It is to do with temporal as well as spatial
correlations. So delayed choice twin slit experiments are a stark demonstration.

The point is that delayed choice twin slit experiments don't
tell us anything more about time than we already knew.

This might be how a human observer chooses to index
events - motivated by a belief in cause and effect or locality. But nonlocality raises the question of when an event really occurs.
The point I was making was that a global scale observer would
see both ends of an event as part of the same effective moment.
So both the photon emission by a star in some distant galaxy and
its "much later" absorbtion by my eye.

An observer whose field of vision includes a supernova a
billion light years from earth and also earth will observe
the light created by the supernova taking a billion years to
reach earth. He wouldn't see the birth of the supernova
and the recording of a picture of it on earth as happening
at the same time.

Questions about when events occur arise when we use different
clocks to index the occurance of spatially separated events.
In a global or 'system' context, which is what 'nonlocality' refers
to, if the spatially separated events are timed by the same
clock then the temporal relationship between the events
is less problematic.

For example, wrt the 'twin paradox' of special relativity,
if a 'global' observer were to time the traveller's journey by
using, say, revolutions of the earth, then he would observe
that the journey took a certain number of revolutions -- which
would be the same for the earthbound twin as for the
travelling twin.

What the proof of nonlocality tells us is that locality-based models of causality are incomplete (though they are certainly still useful). So I was talking about the kinds of models based on hierarchy theory that might present a different view of time.

Nonlocality refers to spatially separated events that are parts
of a single behavioral system. Nonlocality is evident in nature.
There is a hierarchy of systems, or observational contexts.
The scale of behavior/observation doesn't change the basic
meaning of 'time', or contradict the standard notion of
local causality.

Yeah but then you have the problem of the chicken and the
egg. Cause and effect thinking runs into paradoxes - like how
can a something (the universe) spring out of a nothing.

The origin of the universe will remain a matter of untestable
speculation for a long time I think.

Anyway, cause and effect thinking isn't paradoxical. Causes
can't happen after the effects that they cause, by definition.
Given any, causally related, chicken-egg duo, either the egg
was laid by the chicken or the chicken hatched out of the
egg. If they 'sprang' into existence at the same time, then
they're not causally related to each other -- but they might
be nonlocally related as parts of a system that encompasses
them both.

So to get out of this, you have to look into other causal models.
So for example, ones that start with a state of vague everythingness (cf: Anaximander, Peirce) and then dichotomise or symmetry-break to produce two crisp limits on being. So we would now start with a vague chicky-egginess and watch it divide asymmetrically into a chicken and egg
(or if you like, the first egg inside the first chicken). So now causes are effects.

There's nothing to "get out of."

Two events are causally related if there is an invariant,
sequential relationship wrt their occurance. Which one is
called the cause and which one the effect depends on their
relative placement in the temporal indexing of the sequence
of events.

I realise that alternatives to locality and mechanistic logic are not fashionable. But non-locality has to be accounted for within some
causal model unless you want this aspect of reality to remain
a mystery.

What remains to be understood is the deep, qualitative nature of
reality. We can describe/predict the gravitational behavior of
macroscopic objects pretty accurately, but don't know what
causes it. We can predict rates of coincidental detection in
Bell tests pretty accurately, but don't know what's happening
at the level of paired emissions. We can predict detection
patterns in photon/electron two-slit interference experiments,
but don't know what's happening at the level of the emissions
interacting with the two-slit and detection devices.

It's not a matter of reinventing or redefining causality or
time. There just isn't enough data.

 

[mccrone]

Anyway, cause and effect thinking isn't paradoxical. Causes
can't happen after the effects that they cause, by definition.
Given any, causally related, chicken-egg duo, either the egg
was laid by the chicken or the chicken hatched out of the
egg. If they 'sprang' into existence at the same time, then
they're not causally related to each other -- but they might
be nonlocally related as parts of a system that encompasses
them both.

This kind of mutual causality can arise in the most unlikely places - for example, Newton's third law. For every action an equal and opposite reaction. I throw a ball and the ball "throws" me. Both action and reaction spring up at precisely the same moment. And even though you might still want to say one led to the other, put the situation into deep space - say two masses colliding - then you really don't know which exerted the action, which responded with a reaction.

Think then about why this strange law was necessary. Newton made the world mechanical cause and effect with the first two laws. Then had to add back in the deeper mutuality between figure and ground, event and context, as a fictitious symmetric force of reaction.

You don't need to get into QM or relativity to find causal weirdness in mechanical logic - cause and effect thinking.

Again, mechanical logic is a very effective tool in modelling. No surprise that it is first choice when pragmatism rules. But that does not close the door on broader causal models that may capture more of the truth of reality.

What remains to be understood is the deep, qualitative nature of
reality. We can describe/predict the gravitational behavior of
macroscopic objects pretty accurately, but don't know what
causes it. We can predict rates of coincidental detection in
Bell tests pretty accurately, but don't know what's happening
at the level of paired emissions. We can predict detection
patterns in photon/electron two-slit interference experiments,
but don't know what's happening at the level of the emissions
interacting with the two-slit and detection devices.
It's not a matter of reinventing or redefining causality or
time. There just isn't enough data.

I'm not sure I see the logic in your position here. The data tells us our causal models are inadequate. So rather than reconsider our causal models, let's gather more data.

Even if we disregard the evidence that science is theory-led rather than data-led on the whole (ie: what looks like data, what you feel is worth measuring, is determined by what you believe is probably happening), you seem to want to put unnecessary limits on enquiry.

And to return to the specific issue of temporal sequence, are you saying that delayed choice twin slit experiments don't seem to put the cause of a choice of path after the apparent effect, the actual choice of a path?

Cheers - John McCrone.

 

[Sherlock]

I'm not sure I see the logic in your position here. The data tells us our causal models are inadequate. So rather than reconsider our causal models, let's gather more data.

Even if we disregard the evidence that science is theory-led rather than data-led on the whole (ie: what looks like data, what you feel is worth measuring, is determined by what you believe is probably happening), you seem to want to put unnecessary limits on enquiry.

And to return to the specific issue of temporal sequence, are you saying that delayed choice twin slit experiments don't seem to put the cause of a choice of path after the apparent effect, the actual choice of a path?

Cheers - John McCrone.

The delayed choice twin-slit experiments illustrate how what
you see depends on how you look at something. With both
slits open, if you only have two places where a localized photon or
electron detection can appear, then it will appear in one of the
two places. If you use a continuous screen, then eventually
you'll get an interference pattern. In either case, or even if
one slit is closed, there's never any path information, per se.

Must go now, but will discuss more of your comments later
today.

 

[Sherlock]

This kind of mutual causality can arise in the most unlikely places - for example, Newton's third law. For every action an equal and opposite reaction. I throw a ball and the ball "throws" me. Both action and reaction spring up at precisely the same moment. And even though you might still want to say one led to the other, put the situation into deep space - say two masses colliding - then you really don't know which exerted the action, which responded with a reaction.

Did the ball leave your hand before your arm moved?
If so, then the ball leaving your hand would be called the action or
cause and the the movement of your arm would be called the
reaction or effect.

In the case of two bodies colliding in deep space, the collision
event can be said to cause the events that can be
immediately, locally related to it -- such as the fragmenting of,
or changing the trajectories of the colliding bodies.

Think then about why this strange law was necessary. Newton made the world mechanical cause and effect with the first two laws. Then had to add back in the deeper mutuality between figure and ground, event and context, as a fictitious symmetric force of reaction.

You don't need to get into QM or relativity to find causal weirdness in mechanical logic - cause and effect thinking.

Again, mechanical logic is a very effective tool in modelling. No surprise that it is first choice when pragmatism rules. But that does not close the door on broader causal models that may capture more of the truth of reality.

I think Newton probably played a lot of croquet, and observed
that not only did a stationary ball move when hit by a moving
one, but the moving ball slowed down or stopped following
the collision -- and that this behavior (due to interaction) could
be related, quantitatively, in a general way.

There's nothing causally weird about mechanical logic.
It's just a way of talking about things that facilitates
quantification.

I'm not sure what you mean by "broader causal models that
capture more the truth of reality". Nonlocal contexts aren't
causal, they're correlational.

I'm not sure I see the logic in your position here. The data tells us our causal models are inadequate. So rather than reconsider our causal models, let's gather more data.

If you mean, eg., Bell tests, then it's not that "our causal models
are inadequate" in general or anything like that. It's that Bell test
setups are not causal contexts. They're correlational contexts.
Events at A and B are correlated wrt variations in a global variable.
You can trace the chain of events in such experiments and see that
local causality isn't violated (even if there are ftl signals travelling from
A to B or vice versa). There's simply the open question of
how are the actual physical disturbances that are travelling from
emitter to polarizers to detectors related (if they are related) to each
other, and when/where is the relationship (that might be relevant for
the predictable correlations via the global variable) created.
This is an empirical question, I think.

Even if we disregard the evidence that science is theory-led rather than data-led on the whole (ie: what looks like data, what you feel is worth measuring, is determined by what you believe is probably happening), you seem to want to put unnecessary limits on enquiry.

What unnecessary limits?

And to return to the specific issue of temporal sequence, are you saying that delayed choice twin slit experiments don't seem to put the cause of a choice of path after the apparent effect, the actual choice of a path?

The temporal sequence is 1. emission 2. choose detector 3. record
detection.

There isn't any path information. There's just two different
detection methods, one of which is in place *before* the
disturbance or disturbances transmitted by the twin-slit reaches
the detector. With both slits open there's just no way to tell
if something went through both slits or only one.

 

[mccrone]

In the case of two bodies colliding in deep space, the collision event can be said to cause the events that can be
immediately, locally related to it -- such as the fragmenting of,
or changing the trajectories of the colliding bodies.

You're missing the point. Of course the third law is about locality. But even in the classical modelling of locality there is an example of chicken and egg paradox. Newton smuggled in a reaction that arises instantly (not before, not afterwards). Einstein then went on to exploit the in principle impossibility of saying one part of the system did the moving, the other part stayed still.

Nonlocal contexts aren't causal, they're correlational.

You could equally well say that locality is a model and therefore all that is observed are correlations with the model. If there was no problem created by nonlocality, then people (like Einstein) wouldn't have expended so much effort on hidden variable explanations.

It is easy to agree that you don't need to "explain" nonlocality if your concerns are only practical. Modelling the world in a local way captures enough truth for most human purposes. But still, nonlocality exists and is not - by general agreement - explainable by a local logic.

Locality is violated in QM. But that fact isn't visible to a local observer. The question for the modelling of nonlocality is does it make sense to talk of global observers?

Cheers - John McCrone.

 

[Sherlock]

You're missing the point. Of course the third law is about locality. But even in the classical modelling of locality there is an example of chicken and egg paradox. Newton smuggled in a reaction that arises instantly (not before, not afterwards). Einstein then went on to exploit the in principle impossibility of saying one part of the system did the moving, the other part stayed still.



You could equally well say that locality is a model and therefore all that is observed are correlations with the model. If there was no problem created by nonlocality, then people (like Einstein) wouldn't have expended so much effort on hidden variable explanations.

It is easy to agree that you don't need to "explain" nonlocality if your concerns are only practical. Modelling the world in a local way captures enough truth for most human purposes. But still, nonlocality exists and is not - by general agreement - explainable by a local logic.

Locality is violated in QM. But that fact isn't visible to a local observer. The question for the modelling of nonlocality is does it make sense to talk of global observers?

Cheers - John McCrone.

I could be missing the point. Maybe we're using the word,
"nonlocality", differently. I'm thinking of it as just referring
to context. The predictable variable results gotten wrt nonlocal
observational contexts are understood as being due to
looking at spatially separated parts of the same system,
or looking at spatially separated objects that have interacted
or have a common origin.

Of course it makes sense to talk of global observers -- but, I'm
not sure what it means to say that "locality is violated in QM."
I mean, if the observational context is a combined, rather than
an individual, one, then is locality 'violated'?