Demonstrating Social Error Recovery with AgentFactory

Abstract

Exception handling is a well established method of error recovery through the alteration of plans in situ. This method relies on recovery routines existing in advance, which- we argue- is simply shorthand for more detailed plan descrip-tions. However, in practice, agents rarely act alone in their environment- other agents may exist, and potentially pro-vide help in times of need. We argue that social error re-covery is a particular class of exception handling that al-lows agents to resolve erroneous situations that are beyond their direct control. In our opinion, agent oriented program-ming languages must directly provide agents with abilities like social error recovery. Consequently, we introduce re-visions to the AgentFactory framework, and more specifi-cally, the programming language (AF-APL), which facili-tate the rapid development of agents with in-built social er-ror recovery. The use of these abilities are illustrated via an example of a social error recovery scenario for a mo-bile robot working as an office assistant. 1.