Deep Incentive Design with Differentiable Equilibrium Blocks
This work addresses the problem of designing multi-agent interactions with desirable equilibrium outcomes for researchers and practitioners in economics and computer science, offering a unified and scalable approach.
This paper tackles the challenge of automated design of multi-agent interactions with desirable equilibrium outcomes, which is difficult due to computational hardness, non-uniqueness, and instability. The authors propose a differentiable framework called deep incentive design (DID) that uses game-agnostic differentiable equilibrium blocks (DEBs) as modules. This framework is validated on contract design, machine scheduling, and inverse equilibrium problems, handling games with two to sixteen actions per player.
Automated design of multi-agent interactions with desirable equilibrium outcomes is inherently difficult due to the computational hardness, non-uniqueness, and instability of the resulting equilibria. In this work, we propose the use of game-agnostic differentiable equilibrium blocks (DEBs) as modules in a novel, differentiable framework to address a wide variety of incentive design problems from economics and computer science. We call this framework deep incentive design (DID). To validate our approach, we examine three diverse, challenging incentive design tasks: contract design, machine scheduling, and inverse equilibrium problems. For each task, we train a single neural network using a unified pipeline and DEB. This architecture solves the full distribution of problem instances, parameterized by a context, handling all games across a wide range of scales (from two to sixteen actions per player).