# Bell Non Locality, Quantum Non Locality, Weak Locality, CDP

• B
rubi
There are some situations, where the factorization condition can also be violated classically. For example, if you perform post-selection on some data set. Assume Alice and Bob throw dies and the corresponding data sets are ##(A_i)_i## and ##(B_i)_i##. Then you can post-select only those events where ##A_i = B_i## and you will get perfect correlations even though the factorization condition will be violated. There are also some other ways to violate the condition. Classically, all of them can be fixed by coming up with more general conditions, but it already shows that the factorization condition is a heuristic rather than a law of nature.

zonde
Gold Member
I will try to make my point using older statement in this thread:
The question is: Can the EPR argument be applied to the situation when Alice and Bob measure different angles? And the answer is undeniably no, it can't, because in such a situation, Alice would have to make a prediction that cannot even in principle be tested experimentally.
This statement is of course correct but it is missing the point.
So I would ask different question: Can the EPR argument be applied to the model that is Einsten's local and not superdeterministic?
And the answer is yes, using following reasoning:
1) For a model that is not superdeterministic measurement angles are external parameters and it has to produce predictions for any measurement angle.
2) If model is Einsten's local it has to produce predictions independently for Alice and Bob (when Alice's and Bob's measurements are spacelike separated).
Putting 1) and 2) together the model has to produce two sets of independent predictions for Alice and Bob that can be compared and for the cases where Bob's and Alice's measurement angles are the same we can apply EPR argument.

So the next question would be - can we compare (correlate) predictions of model where Alice's and Bob's measurement angles are different? And the answer to this question seems to be that they better be comparable as we do that a lot in real entanglement experiments.

rubi
This statement is of course correct but it is missing the point.
No, it doesn't miss the point. You don't want to acknowledge the fact that unmeasurable quantities don't need to exist and that models needn't model unmeasurable quantities. And there are examples that prove that you are wrong.

For a model that is not superdeterministic measurement angles are external parameters and it has to produce predictions for any measurement angle.
That's just false. You have just stated your personal belief without any argument. You have hidden variables in mind and think that your intuitions about them also hold for non-hidden variable models. But there is just no way to argue that unmeasurable quantities must exist.

What you and Denis are trying to argue is: There must be predictions for any angle and thus we can use the EPR argument to conclude that there must be predictions for any angle. It's circular reasoning and cannot be saved. Either you postulate it as an axiom, as you just did. Then it can be denied (for instance by QM). Or you try to use the EPR argument, but then you must admit that your reasoning is circular and therefore unacceptable.

So the next question would be - can we compare (correlate) predictions of model where Alice's and Bob's measurement angles are different? And the answer to this question seems to be that they better be comparable as we do that a lot in real entanglement experiments.
We don't correlate quantities that never co-occur, such Bob's spins along different angles. It doesn't even make sense to speak about correlations of things don't co-occur. QM beautifully prevents us from doing this by having the corresponding operators not commute. The correlations between Alice and Bob can of course also be calculated in QM, but that's irrelevant for your argument, since correlations for different angles are not perfect and thus can't be used to predict anything with certainty, contrary to what the EPR argument would require.

Last edited:
zonde
Gold Member
You don't want to acknowledge the fact that unmeasurable quantities don't need to exist and that models needn't model unmeasurable quantities.
I acknowledge that unmeasurable quantities don't need to exist and that models needn't model unmeasurable quantities.
For a model that is not superdeterministic measurement angles are external parameters and it has to produce predictions for any measurement angle.
That's just false.
Which part is false?
Is this part false? - "For a model that is not superdeterministic measurement angles are external parameters"
Or the other part?
Argument for the other part (if the first part is ok) is as follows:
As measurement angles are external parameters, experimentalist can choose whichever angle he wants and test prediction for that angle. Predictions have to be made before test is performed.
As experimentalist's choice lies outside the model, predictions have to be made independently from that choice and before the measurement.
For me it seems enough to claim that the model should be capable of producing predictions for any measurement angle.

rubi
Is this part false? - "For a model that is not superdeterministic measurement angles are external parameters"
Yes, this part is false. QT is capable of modeling measurement angles within the model and it is definitely not superdeterministic. No fine-tuning is required.

Or the other part?
Argument for the other part (if the first part is ok) is as follows:
As measurement angles are external parameters, experimentalist can choose whichever angle he wants and test prediction for that angle. Predictions have to be made before test is performed.
As experimentalist's choice lies outside the model, predictions have to be made independently from that choice and before the measurement.
For me it seems enough to claim that the model should be capable of producing predictions for any measurement angle.
This is false as well. An experimenter can predict whatever he or she wants. This does not imply that there must be something corresponding to that prediction. If I predict that there is a pink unicorn behind you, it is not necessarily true. If Alice predicts that the spin of Bob's particle along the angle ##\alpha## is so and so, even though his detector is aligned along a different angle ##\beta\neq\alpha##, this doesn't imply that Bob's particle has a spin along the angle ##\alpha##. And given that it is impossible in principle to test such a prediction, there is no reason to expect that the prediction would be correct. And we understand the issue very well. QT is contextual and that means that properties that don't commute with all observables emerge from the experimental setup rather than existing independent of the setup.

zonde
Gold Member
Yes, this part is false. QT is capable of modeling measurement angles within the model and it is definitely not superdeterministic. No fine-tuning is required.
Well, but you can model only one measurement angle at the same time, and which particular angle you are modeling you take from outside the model as external parameter.
Maybe this will be more clear - "For a model that is not superdeterministic choice of measurement angle is external parameter"
This is false as well. An experimenter can predict whatever he or she wants. This does not imply that there must be something corresponding to that prediction. If I predict that there is a pink unicorn behind you, it is not necessarily true.
"Pink unicorn behind you" is hypothesis not prediction. Prediction would be statement about what I would observe if I turn around.
If Alice predicts that the spin of Bob's particle along the angle ##\alpha## is so and so, even though his detector is aligned along a different angle ##\beta\neq\alpha##, this doesn't imply that Bob's particle has a spin along the angle ##\alpha##.
Predictions are conditional. If you do such and such you will observe this. What you are talking about is hypothesis not prediction.