Social Optimization in the Presence of Cognitive Local Optima: Effects of Social Network Topology and Interaction Mode

The combined tendency for positive self-presentation and mimicry or social learning results in the capability of a population of simulated individuals to optimize their cognitive structures. A population of parallel constraint satisfaction networks was created, with globally and locally optimal activation patterns. Individuals started with random activations and interacted to find optimal vectors. Two social network topologies were tested, as well as two modes of interaction; results indicated that the ability to optimize activation vectors depends on the social network configuration and whether individuals are influenced by their best neighbor or are attracted to a neighborhood centroid. Simulated individuals are able to find suitable patterns of belief or attitude with very little internal information processing, using social interaction.