Probationary Contracts: Reducing Risk in Norm-Based Systems

Abstract In human organisations, it is common to subject a new employees to periods of probation for which additional restrictions or oversight apply in order to reduce the consequences of poor recruitment choice. In a similar way, multi-agent organisations may need to employ agents of unknown trustworthiness to perform services defined by contracts (or sets of norms), yet these agents may violate the norms for their own advantage. Here, the risk of employing such agents depends on the agents trustworthiness and the consequences of norm violation. In response, in this paper we propose the use of probationary contracts, generated by adding obligations to standard contracts in order to further constrain agent behaviour. We evaluate our work using agent-based simulations of abstract tasks, and present results showing that using probationary roles reduces the risk of using unknown agents, especially where violating a norm has serious consequences.

Authors

• Chris Haynes
• Simon Miles
• Michael Luck

Keywords

No keywords are indexed for this paper.