Agent Incentives: A Causal Perspective
Abstract
We present a framework for analysing agent incentives using causal influence diagrams. We establish that a well-known criterion for value of information is complete. We propose a new graphical criterion for value of control, establishing its soundness and completeness. We also introduce two new concepts for incentive analysis: response incentives indicate which changes in the environment affect an optimal decision, while instrumental control incentives establish whether an agent can influence its utility via a variable X. For both new concepts, we provide sound and complete graphical criteria. We show by example how these results can help with evaluating the safety and fairness of an AI system.
Cite
@article{arxiv.2102.01685,
title = {Agent Incentives: A Causal Perspective},
author = {Tom Everitt and Ryan Carey and Eric Langlois and Pedro A Ortega and Shane Legg},
journal= {arXiv preprint arXiv:2102.01685},
year = {2021}
}
Comments
In Proceedings of the AAAI 2021 Conference. Supersedes arXiv:1902.09980, arXiv:2001.07118