Automata Simulation of N-Person Social Dilemma Games.
Vladimir Akimov, Mikhail Soutchanski
Abstract.
Collective behavior of N players in a social dilemma game is
simulated by automata exhibiting asymptotically cooperative
behavior. In his automata models of simple biological systems
M.Tsetlin assumed minimum of information available to the
"players." Our automata were somewhat more sophisticated, using
Markov strategies in their interactions. We investigated
relationships between information received by the automata and
the emergence of cooperation in a simulated evolution process. In
some ways our approach is similar to that of R.Axelrod. It
differs in that instead of determining the "most successful"
strategy, we seek surviving strategies in a social dilemma
environment. Previous result by M.Soutchanski showed that
cooperation could be established asymptotically under partially
centralized control. In our model there is no such control. Our
main result is that more sophisticated behavior of "self-seeking"
automata compensates for the absence of such control. Moreover
cooperation is established more rapidly when more information is
available to the automata.
Axelrod R. "The evolution of cooperation". Basic Books, 1984, N.Y.
Soutchanski M.E. "Adaptive algorithm for determination of
weakly efficient variant under randomness". In: Soviet J. of
Computer and Systems Sciences, 1987, v.25, N3, p.148-157
(name of the author was translated as Suchanskiy M.E.)
Tsetlin M.L. "Automaton theory and modeling of biological
systems". (ser. Mathematics in Science and Engineering, v.
102), Academic Press, N.Y., 1973, 288p.