Philosophical Problems of Multi-Agent Systems Modeling

Abstract

Multi-agent systems (MAS) is a technology based on distributed computing, designed to emulate the complex emergent behavior of multiple agents on the basis of various feedback mechanisms. Philosophical interest therein is caused, firstly, by the cognitive properties of individual agents and of the system as a whole; secondly, with the possible modeling of complex social behavior and, possibly, by the perspective of a progress in social understanding. The main problem is the extent to which multi-agent models correspond to the simulated reality. Since they are computing systems, it is necessary to consider the general theory of computations and their ontological status; to determine how the computational processes form rule-like systems consisting of multiple cognitively enabled agents; to make plausible assumptions about the regularities of the evolution of such systems under natural and artificial conditions. Particular attention is paid to the hypothesis of the limited rationality of agents or, what is the same, to their cognitive limitations, which demands the application of epistemic logic formalism to MAS functioning. The obtained results give additional arguments in favor of the well-known philosophical thesis: the higher cognitive capabilities are determined and explained by the complex social organization of their bearers.