One has to be very careful not to underestimate the intellectual capabilities
of random numbers. Human beings tend to refer to them in denigrating terms
as "noise" and this may be just the common cultural phenomenon of casting
perceived superiors as inferiors through pejorative verbal labels.
Ross Ashby showed that random numbers exhibit many psychological phenomena
such as habituation. Rabin showed that random automata can recognize
languages that are beyond the capabilities of Turing machines. I showed
that there are practical pattern recognition and control problems that
are insoluble by any finite state automaton, require very complex
recursive functions to solve, and yet are trivially soluble by zero-memory
stochastic automata.
In human-computer interaction a degree of randomness in the choice of
computer responses makes the system appear very much more intelligent --
this works well for humans also.
I think, in these terms, the results you describe are unsurprising, and Bob
has no explaining to do.
b.
Dr Brian R Gaines Knowledge Science Institute
University of Calgary
gaines@cpsc.ucalgary.ca Calgary, Alberta, Canada T2N 1N4
403-220-5901 Fax:403-284-4707 http://ksi.cpsc.ucalgary.ca/KSI
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%