Robust trust: Prior knowledge, time and context

Trust is an agent's expectation of the value it will observe when it evaluates the enactment of another agent's commitment. There are two steps involved in trust: first the action that another agent is expected to enact given that it has made a commitment, and second the expected valuation of that action when the result of that action is eventually consumed. A computational model of trust is presented that takes account of: prior knowledge of other agents, the evolution of trust estimates in time, and the evolution of trust estimates in response to changes in context. This model is founded on the principle of information-based agency that each and every utterance made contains valuable information. The computational basis for the model is substantially simpler and is more theoretically grounded than previously reported. © 2012 Springer-Verlag