Deceiving in GOLEM: how to strategically pilfer help

We first introduce GOLEM, a Multi-Agent world designed to study: different social attitudes and personalities in interaction and, in particular, attitudes in delegating tasks to other agents (search for help) and in adopting goals of others (help); reasoning about delegating and adopting tasks and about the other’s personality. Secondly, we shortly present the different forms of deception that are allowed in this world. This is also useful to provide some basic ontology of deception and to clarify the relationships between deception and lie, deception and mistake and so on. Third, in the main part of the paper, we focus on one family of examples of deception in GOLEM, which are particularly relevant: social and intentional deceptions, about the agent itself (its personality, its capabilities, its goals) rather than about the world, and in order to obtain help (instrumental deception), which is (till now) the most important social motive for GOLEM agents; all these forms of deception are not necessarily based on implicit or explicit communication (lie). Finally, we consider some more advanced forms of lie that cannot be treated in the current implementation of GOLEM but are very interesting for their strategic character: in these forms of lie, the deceiving agent predicts and exploits the reasoning that the deceived agent will perform as a consequence of her (false) speech acts (request, refusal, etc.). All this just in order to: (i) clarify important basic problems of interaction among cognitive agents; (ii) show how even a toy world and a simple agent platform may be theoretically productive; (iii) point out interesting actions and reasoning to be formalized and implemented.