Active Inference via Incentive Design: Eliciting Divergent Behaviors under Partial Observability
The paper introduces a method for active inference using incentive design, framed as a leader-follower game, to infer an unknown follower's type from partial observations in a stochastic system. The leader strategically designs side payments (incentives) to influence the follower's behavior, balancing the gain in information about the follower's type against the cost of these incentives.