Reciprocating with Learned Models

Anish Biswas and Sandip Sen

In our prior work, we have demonstrated the effectiveness of a probabilistic reciprocity mechanism by which self-interested agents may learn to adopt a cooperative relationship with other similar agents. Reciprocative decisions were made based on the balance of help with another agent. In this paper, we expand that framework by making an agent explicitly model the help-giving procedure of the other agent. This learned model is then used to make decisions on requests for help received from that agent. We show that when asking for help consumes non-negligible time, the model based reciprocative agents can outperform the reciprocative agents who uses only balance of past transactions.

This page is copyrighted by AAAI. All rights reserved. Your use of this site constitutes acceptance of all of AAAI's terms and conditions and privacy policy.