Optimizing Mechanism Design in Multi-Agent Reinforcement Learning
Bibliographic record
Abstract
The increasing adoption of agentic systems for decentralized deployment of artificial intelligence presents new challenges in efficiently selecting parameters that influence agents' learned behavior, applying mechanism design in complex settings, and combining diverse agent capabilities to achieve desired outcomes. Algorithmic mechanism design, an interdisciplinary field bridging computer science, mathematics, and economics, develops algorithms that guide rational agents toward desired behaviors. Its applications include resource allocation, cost-sharing, pricing, and combinatorial auctions. However, traditional methods face limitations due to computational constraints and static assumptions, making them less effective in dynamic environments characterized by uncertainty and change. This thesis addresses these limitations by integrating reinforcement learning (RL) and Bayesian Optimization (BO) to develop adaptive mechanisms in dynamic multi-agent settings. We introduce new frameworks for mechanism design in Multi-Agent Reinforcement Learning (MARL), relying on novel BO-based approaches to efficiently explore promising designs. MARL captures the complexity of dynamic interactions among multiple agents in stochastic environments, solving the underlying Markov Game to learn a joint policy. The computational complexity of evaluating multiple MARL scenarios is addressed by (I) extending Successor Features for transfer learning to Nash Equilibrium policies and (II) using BO-based frameworks to limit the evaluation budget, making the problem tractable. The effectiveness of the proposed mechanism design frameworks is showcased through extensive benchmark studies in practical applications such as setting service charges for drivers in taxi platforms, managing the exploitation of shared natural resources to maximize social welfare, optimizing hardware acquisition decisions in robot fleets for exploration missions, and defining optimal incentive and recruitment policies to maximize principal's objectives. Demonstrating the superiority of our methods over the state-of-the-art in real-world problems highlights the potential of integrating BO and MARL to optimize complex multi-agent systems, providing a robust foundation for future research in mechanism design.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.006 | 0.013 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".