On Various Sources of Uncertainty in Modeling Suspicion and How to Treat Them

Among the several meanings that trust may take, ‘trust about correct information’ is strictly related to some information source properties such as sincerity and cooperativity (Demolombe, in press) and to the information receiver’s arising of suspicion, about the information source and about the communicated data. To defend themself from deception, fraud and other forms of trust-violating conditions, multiagent systems may adopt at least two approaches: they may establish norms and rules, and sanctions in case these are violated, or may set up mechanisms to simulate arising of suspicion in interactions. The two approaches are not alternative (as humans know very well), and each of them has its own advantages and disadvantages. The main advantage of establishing norms and controls rather than trying to detect deception is that ‘expectation of truth’ may be assumed as a default condition, and interaction may therefore be simplified. Keeping an almost permanent attitude of suspicion, on the contrary, risks to delay message interpretation beyond the acceptable: in addition, discovery of deception requires a rather deep knowledge of the mental state of the message Sender (from the Receiver), which violates the ‘open world’ hypothesis that is typical of multiagent systems.