Research map / Multi-agent systems and games / Cooperative multi-agent RL
AI for Global Climate Cooperation: Modeling Global Climate Negotiations, Agreements, and Long-Term Cooperation in RICE-N
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Introduces RICE-N, a MARL-based multi-region integrated assessment model with negotiation protocols (Bilateral Negotiation, Basic Club) to study climate cooperation and its effect on temperature, emissions, output and inequality.
Problem
Climate mitigation requires cooperation among strategic regions without central enforcement, but existing integrated assessment models (RICE/DICE) use exogenous fixed policies and lack strategic negotiation dynamics and distributional analysis.
Approach
Embed negotiation protocols (bilateral proposals and a climate-club tariff mechanism inspired by Nordhaus and the EU CBAM) into a calibrated RICE-based economic-climate simulator where regional agents learn policies endogenously via multi-agent reinforcement learning (A2C), rather than using hand-set policies.
Key results
- Basic Club and Bilateral Negotiation reduce 2115 temperature anomaly to 3.42 and 3.34 C versus 5.12 C with no negotiation.
- Gini index for abatement cost falls from 0.695 (no negotiation) to 0.339 (Basic Club) and 0.333 (Bilateral).
- Basic Club yields 2115 global output of 534.7 trillion USD versus 1697.2 for no negotiation? (table reports carbon 534.7 and consumption 8.55 trillion USD, with Basic Club ending steady output growth).
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use Basic Club instead of no negotiation | RICE-N agents · RICE-N year 2115 | temperature anomaly: 5.121±0.032 -> 3.422±0.018 | Helped |
| use Bilateral Negotiation instead of no negotiation | RICE-N agents · RICE-N year 2115 | carbon emissions: 1697.2±25.6 -> 487.0±15.2 | Helped |
| use Basic Club instead of no negotiation | RICE-N agents · RICE-N year 2115 | global output: 233.6±3.9 -> 235.1±5.0 | No clear effect |
| use Bilateral Negotiation instead of no negotiation | RICE-N agents · RICE-N year 2115 | global output: 233.6±3.9 -> 209.4±5.5 | Hurt |
| use Basic Club instead of no negotiation | RICE-N agents · RICE-N year 2115 | abatement cost Gini: 0.695±0.018 -> 0.339±0.006 | Helped |
| use Bilateral Negotiation instead of no negotiation | RICE-N agents · RICE-N year 2115 | mitigation rate Gini: 0.405±0.018 -> 0.011±0.007 | Helped |
| use Basic Club instead of Bilateral Negotiation | RICE-N agents · RICE-N year 2115 | global output: 209.4±5.5 -> 235.1±5.0 | Helped |
Surprising findings
- Bilateral Negotiation achieves stronger climate outcomes than Basic Club, attributed to the max-of-proposals commitment mechanism producing near-maximal mitigation early in training.
- No negotiation yields the highest consumption in the short run but its economic advantage is expected to reverse as warming damages accumulate.
Limitations
- No regional damages or temperatures modeled; damage disparities within regions ignored.
- Negotiation reasoning is a black box; binding commitments assumed with no deviation within a 5-year step.
- Bilateral Negotiation is an illustrative protocol with O(n^2) communication and limited real-world applicability.
Experimental setup
Datasets: DICE2016 calibration parameters Howard & Sterner damage meta-analysis FaIR climate model
Benchmarks: RICE-N simulated climate-economic outcomes
Metrics: global temperature anomaly carbon emissions global output production global consumption Gini index
Baselines: No negotiation Minimum mitigation Maximum mitigation
Builds on: RICE DICE2016 A2C AI Economist Armington trade model
Closest papers by meaning
- InvestESG: A multi-agent reinforcement learning benchmark for studying climate investment as a social dilemma (ICLR 2025 · similarity 0.86)
- Cooperate or Collapse: Emergence of Sustainable Cooperation in a Society of LLM Agents (NeurIPS 2024 · similarity 0.84)
- Revisiting Multi-Agent World Modeling from a Diffusion-Inspired Perspective (NeurIPS 2025 · similarity 0.83)
- Melting Pot Contest: Charting the Future of Generalized Cooperative Intelligence (NeurIPS 2024 · similarity 0.83)
- Cooperative Policy Agreement: Learning Diverse Policy for Offline MARL (AAAI 2025 · similarity 0.82)
- Optimizing Heat Alert Issuance with Reinforcement Learning (AAAI 2025 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.