Dear Mauro and Peter,
I read your exchange with interest and apologize for not replying sooner -- I've been traveling (a combination of science and music), so things got put off.
I agree that one should not try to use chaos as the "little monster" that can explain everything, especially at this early stage, when things are pretty much empirical and by analogy. Actually, the violations of Bell's inequalities are more related to Bayesian statistics than they are to chaos. The point of attack is on the classical side, for classical nonlinear systems can exhibit correlations (analogous to entanglement) essentially as large as those in quantum mechanics. Thus, the violations of the inequalities are ruling out the lack of correlations (in linear systems?!) rather than classical mechanics per se. (If you look back at the so-called "classical" derivation of, say, the CHSH inequality, which is the most experimentally friendly version and the one commonly used, you will find that there are really no correlations built in -- they are just glossed over, whereas the on the quantum mechanical side one normally starts with a singlet state, which as about as entangled as you can get.) These classical correlations have been studied extensively in systems as diverse as tornados and energy distributions of cosmic rays, and they exhibit so-called nonextensive (Tsallis) entropy. The book "Nonextensive Entropy: Interdisciplinary Applications," put together by Gell-Mann and Tsallis (Ref. [7] in my essay) covers this in a relatively straightforward fashion. Currently, the whole business is mostly experimentally driven, so mathematical derivations are at a minimum, but they have had surprising success with quite diverse systems.
A final word. Perhaps I am naive, but coming from an experimentalist's perspective, I find it odd the way people jump from statistical correlations to individual cases. All of the Bell-type inequalities rely on correlations found in large numbers of data, when "enough events have been recorded to be statistically significant and meaningful." (This is true even for the three-state GHZ correlations, which don't rely on an inequality.) Yet, when it comes to the interpretation, people say such things as, "When the spin direction of Alice's particle is measured as 'up,' this has caused its wave-function (previously assumed to be in entangled limbo) to collapse, and since it came from a singlet state, this causes the wave-function of Bob's particle to collapse INSTANTANEOUSLY into a 'down' state." Experimentally, there has been no analysis of individual particle-particle data -- the correlations are meaningful only after many thousands of events have been collected and compared statistically. Surely this is a weakness in the argument!
Again, thanks for the dialog.
Bill