Agent reasoning for norm compliance: a semantic approach
M. Birna van Riemsdijk, Louise A. Dennis, Michael Fisher, Koen V. Hindriks · 2013
A system of autonomous agents may exhibit undesirable or ineffec-tive behavior if no form of regulation is imposed. Norms, describ-ing how agents should ideally behave, can be used to address this issue if agents are able to reason about norms and adapt their behav-ior to comply with them (if they choose to do so). Assuming that which norms will have to be followed is unknown at design time, it is not possible to pre-program agents such that their behavior is norm compliant. Instead, we need a generic execution mechanism that allows agents to adapt their behavior at run-time, which is what we propose in this paper. The execution mechanism is defined on top of an abstract agent decision making mechanism. This is done by allowing the execution of actions by the agent decision mak-ing mechanism only if these are not forbidden according to norms, as well as triggering the execution of actions if this is required by norms. We specify norms using Linear Temporal Logic and define the operational semantics of the execution mechanism using tech-niques from executable temporal logic. We formally analyze prop-erties of the execution mechanism, including norm compliance.