SCAMP : a Stigmergic Approach to Modeling Intelligent Behavior

SCAMP (Social Causality with Agents using Multiple Perspectives) is a social simulator that incorporates a realistic array of psychological phenomena. Its central assumption is that the major causal forces in a social scenario are the choices made by the participants, which include recognizing a dynamically changing array of accessible options and selecting among them, based on actor preferences over the features of options (tactical choice), longer range desires or goals (strategic choice), and thinking into the future (mental simulation), under the influence of social interactions. All of these features are classicallyassociated with intelligence, and a computational system that instantiates themis considered to manifest Artificial Intelligence. Drawing on Herbert Simon’s insight about an ant on a beach, SCAMP shows how such behavior can result, not from internal symbolic reasoning, Bayesian inference, or neural networks, but from stigmergic dynamics among ant-like agents in an external environment that encodes the desired causal constraints. This talk will explain the origins of SCAMP, describe its causal language, illustrate the behaviors it generates, and speculate about its implications for future research.