Modeling Adaptative Social Behavior in Collective Problem Solving Algorithms

Abstract

Collective problem solving can lead to the development of new methods and algorithms that can potentially contribute to novel Artificial Intelligence applications and tools. Socially-inspired optimization algorithms are a class of algorithms that aim at conducting a search over a large solution space using mechanisms similar to how humans solve problems in a social context. Several such algorithms exist in the literature, including adaptations of classical ones, such as Genetic Algorithms. These models, however, do not take into account a fundamental concept in human social systems: the individual ability to adapt problem-solving strategies as a function of the social context. In this paper, we propose and investigate an extension inside a socially-inspired model of collective problem solving which allows one to model agents with such adaptability. This extension is based on the concept of humans as ``motivated tacticians'' and it dictates how agents are to adapt their search heuristics according to their respective social context. We show how this rule can speed up the system's convergence to good solutions and improve the search space exploration. The results contribute towards the design of socially inspired computational systems for collective problem-solving.