Eroding trust : an agent based model to explore how trust flows

The emergence and deterioration of trust between agents is a critical feature of many open distributed systems. This paper reports on a simulation of one aspect of relationships: the behaviour of peer groups when confronted with lying. It presents an abstract model of shared cognitive space, using NetLogo, where agents (1) determine a perception of trust towards other agents, (2) interrogate nearby agents, and (3) reassess their perceptions of trust based on judgements of others. (4) Some agents are discovered to be liars, (5) causing agents to move away from the liar in their cognitive space. Group cohesion is tested by varying (i) decay of agents' memory, (ii) penalty for lying, (iii) proportion of liars, (iv) probability of discovery, and (v) distance an agents moves away from a liar in cognitive space. We find that penalty for lying has little effect on group cohesion, but the other factors can cause significant disruption.