Catálogo de publicaciones - libros

Both learning and reasoning are important aspects of intelligence. However they are rarely integrated within a single agent. Here it is suggested that imprecise learning and crisp reasoning may be coherently combined via the cognitive context. The identification of the current context is done using an imprecise learning mechanism, whilst the contents of a context are crisp models that may be usefully reasoned about. This also helps deal with situations of logical under- and over-determination because the scope of the context can be adjusted to include more or less knowledge into the reasoning process. An example model is exhibited where an agent learns and acts in an artificial stock market.

This paper addresses in multiagent-based simulation (MABS) to explore agents who can reproduce human-like behaviors in the , which is more difficult to be reproduced than in the ( one time bargaining game). For this purpose, we focus on the Roth’s learning agents who can reproduce human-like behaviors in several simple examples including the ultimate game, and compare simulation results of Roth’s learning agents and Q-learning agents in the sequential bargaining game. Intensive simulations have revealed the following implications: (1) Roth’s basic and three parameter reinforcement learning agents with any type of three action selections (, -greed, roulette, and Boltzmann distribution selections) can neither learn consistent behaviors nor acquire sequential negotiation in sequential bargaining game; and (2) Q-learning agents with any type of three action selections, on the other hand, can learn consistent behaviors and acquire sequential negotiation in the same game. However, Q-learning agents cannot reproduce the decreasing trend found in subject experiments.

This paper refines a previously introduced procedure to quantify objective dependence relations between agents of a multiagent system. The quantification of the dependence relations is performed on a specially defined form of reduced dependence graphs, called . The paper also shows how the procedure can be used to determine a measure of the dependence that a society as a whole has on each agent that participates in it and, correlatively, a measure of the negotiation powers of the agents of such society. The procedure is also extended to allow for the refinement of the objective degrees of dependence into subjective ones, through the use of auxiliary coefficients that can represent some subjective aspects of the dependence relationships. A sample calculation of objective degrees of dependence and negotiation powers of agents of a simple multiagent system is presented, and a hint is given on how degrees of dependence could be used to support social reasoning processes.