Random number generators in Psi experiments (07 Nov 2010)

(Note: I'm not making any comment on the research in question, merely making a technical remark.)

In Feeling the Future, the author, D. Bem, has an interesting problem (page 11 of the linked preprint). If an experiment shows that people can predict the output of a RNG, what does that tell us?

Well, if the RNG was a PRNG, seeded at the beginning of the experiment, then the subject could either be showing precognition (i.e. they can sense the result from the future) or clairvoyance (i.e. they can sense the state of the PRNG and thus know the next result). If you clock a true, hardware RNG before the trial then the same possibilities arise.

If you clock a true RNG after the trial however, then the possibilities are precognition and psycokinesis (i.e. the subject altered the state of the RNG, causing the result).

My observation was that you could XOR a PRNG and a RNG. (The latter being clocked after the trial.) In this case, prediction implies, at least, either both clairvoyance (for the PRNG) and psychokinesis (for the RNG), or precognition. That appears to be a useful trick.

However, this rather relies on the result that the sum of two RNGs is at least as random as the best of them. However, if we have a result which suggests precognition, then that doesn't hold! An evil RNG could, when summed with a true RNG, predict its future output and cancel it out, resulting in a less random result!