Moduleco 2 - Concepts of the multi-agent's dynamics 2.2. Individual learning and cognitive hierarchy of the agent's behaviour (a)

P. Bourgine [1996] « Models of autonomous agents and of their coevolutionary interactions »
B. Walliser [1997] « a spectrum of equilibration processes in game theory »

P. Bourgine [1996] ordered agents by way of their increasing cognitive capacities :

Walliser [1997] identified four dynamic processes, which converge towards an equilibrium in a game-theoretical framework. Like Bourgine's typology it is possible to order this process by  way of the agent's increasing cognitive capacities :

• In (the current interpretation of) an EVOLUTIONARY process, each agent  has a Fixed Strategy, and the reproduction of such strategy is proportional to the payoff obtained through agent interactions.
• In BEHAVIORAL Learning, each player modifies his strategies according to the observed payoff obtained from his past actions.
• In EPISTEMIC learning, each player updates his beliefs about others' future actions, according to their actions observed.
• In an EDUCTIVE process, each player has enough information to perfectly simulate others behavior and immediately reaches equilibrium.

Taking agents heterogeneous with respect to their cognitive capacities leads to a coevolutionay process
(for instance : Banerjee, Weibull [1995] "Evolutionary selection and rational behavior in Kirman, Salmon" [1995])

Denis.Phan@enst-bretagne.fr ;  Antoine Beugnard@enst-bretagne.fr