We consider the problem of using logged data to make predictions about what would happen if we changed the `rules of the game' in a multi-agent system. This task is difficult because in many cases we observe actions individuals take but not their private information or their full reward functions. In addition, agents are strategic, so when the rules change, they will also change their actions.

Reviewers found this paper to be an original and useful addition to the field of multi-agent games. While some of the presentation could be clarified (see specific reviewer comments, e.g. about the revelation game), there was a consensus that the paper is generally well-written and clear enough for publication, with the proposed corrections.

This problem arises in a number of mechanism design contexts, where intervening on a system constitutes changing the rules of the game. Calculating counterfactual value requires reasoning about how rule changes affect equilibrium behavior of the agents. Under strong assumptions this counterfactual value is point-identified, but these assumptions are often implausible. The authors present a scheme for relaxing these assumptions, and characterizing the set of values that are compatible with the observed data under this relaxation. The relaxation of point-identification assumptions is presented in terms of a second game, which the authors call the Revelation Game.

We consider the problem of using logged data to make predictions about what would happen if we changed the rules of the game' in a multi-agent system. This task is difficult because in many cases we observe actions individuals take but not their private information or their full reward functions. In addition, agents are strategic, so when the rules change, they will also change their actions. They make counterfactual predictions by using observed actions to learn the underlying utility function (a.k.a. This approach imposes heavy assumptions such as the rationality of the agents being observed and a correct model of the environment and agents' utility functions.

We consider the problem of using logged data to make predictions about what would happen if we changed the rules of the game' in a multi-agent system. This task is difficult because in many cases we observe actions individuals take but not their private information or their full reward functions. In addition, agents are strategic, so when the rules change, they will also change their actions. They make counterfactual predictions by using observed actions to learn the underlying utility function (a.k.a. This approach imposes heavy assumptions such as the rationality of the agents being observed and a correct model of the environment and agents' utility functions.