A community-maintained collection of resources for economists applying microeconomic theory, game theory, econometrics, industrial organization, and behavioral economics to the development of AI systems: post-training, alignment, safety, mechanism design for AI, and multi-agent systems.
Contributions are welcome. If you know of a resource that should be listed here, please open an issue or submit a pull request.
This guide grew out of Kireyev and Maura-Rivero (2026), "From Microeconomics to AI Research: A Guide for Economists".
- Ouyang et al. (2022) -- InstructGPT. The foundational RLHF paper.
- Rafailov et al. (2023) -- DPO. Collapses reward modeling and policy optimization into a single objective.
- Christiano et al. (2017) -- Deep RL from Human Preferences. The original deep RLHF paper; foundational for understanding how human preferences train policies.
- Bai et al. (2022) -- Constitutional AI. How to use AI-generated feedback (RLAIF) instead of or in addition to human feedback.
- Lambert et al. (2024) -- Tulu 3. Introduces reinforcement learning from verifiable rewards (RLVR) as a post-training method using deterministic reward signals.
- DeepSeek-AI (2025) -- DeepSeek-R1. Demonstrates RLVR at scale, producing strong reasoning capabilities.
- Vaswani et al. (2017) -- Attention Is All You Need. The transformer architecture.
- UK AI Safety Institute (2025) -- Economic Theory and Game Theory for AI Alignment. Research agenda with concrete open problems for economists.
- Conitzer et al. (2024) -- Social choice should guide AI alignment. Maps RLHF to social choice theory.
- Dafoe et al. (2021) -- Cooperative AI. The case for game theory in multi-agent AI.
- Igami (2020) -- AI as structural estimation. Translates RL into structural econometric language, making the conceptual bridge between dynamic programming in IO and RL explicit.
- Rahwan, Parkes et al. (2019) -- Machine Behaviour. Argues for studying AI systems as behavioural agents using tools from the social sciences.
- Athey & Luca (2019) -- Economists (and Economics) in Tech Companies. How economists came to play a central role in tech; context for the next frontier in AI labs.
- Breiman (2001) -- Statistical Modeling: The Two Cultures. Foundational essay framing the data-modeling vs. algorithmic-modeling divide that runs between econometrics and ML.
- Sutton (2019) -- The Bitter Lesson. Influential ML manifesto: general methods that leverage compute outperform hand-engineered structure.
- Varian (2014) -- Big Data: New Tricks for Econometrics. First major JEP piece introducing the ML toolkit to economists.
- Mullainathan & Spiess (2017) -- Machine Learning: An Applied Econometric Approach. Reframes the divide as prediction vs. estimation; the sharpest introduction for economists.
- Athey & Imbens (2019) -- Machine Learning Methods That Economists Should Know About. The standard bridge survey.
- Kaplan et al. (2020) -- Scaling Laws for Neural Language Models. Original power-law forms in parameters, data, and compute.
- Hoffmann et al. (2022) -- Training Compute-Optimal LLMs (Chinchilla). Compute-optimal allocation between parameters and data.
- Snell et al. (2024) -- Scaling LLM Test-Time Compute. Test-time compute can substitute for parameters (ICLR 2025).
- Gao, Schulman & Hilton (2023) -- Scaling Laws for Reward Model Overoptimization. The cleanest scaling result for RLHF specifically (ICML 2023).
| Resource | Topic | Link |
|---|---|---|
| Andrew Ng's Machine Learning | Supervised learning, regularization, evaluation | Coursera |
| MIT 6.S191: Introduction to Deep Learning | Neural architectures, training dynamics, updated annually | introtodeeplearning.com |
| DeepMind RL Course (UCL) | Reinforcement learning foundations | YouTube |
| Andrej Karpathy: Neural Networks Zero to Hero | Transformers, attention, tokenization, full training pipeline | YouTube |
Hugging Face Transformers + trl |
Standard library for working with pretrained LLMs; trl provides RLHF/DPO fine-tuning pipelines |
huggingface.co |
| Sutton & Barto: Reinforcement Learning: An Introduction | The standard RL textbook, freely available online | incompleteideas.net |
| LeetCode | Algorithmic interview preparation (non-negotiable for industry roles) | leetcode.com |
| Community | Description | Link |
|---|---|---|
| Cooperative AI Foundation | Research and events on game theory and multi-agent AI. YouTube channel connects economic theory to AI. | cooperativeai.com |
| UK AI Safety Institute | Research agenda on economic theory and game theory for AI alignment, with concrete open problems for economists. | Alignment Project |
| ACM EC community | The annual ACM Conference on Economics and Computation. Natural home for economists publishing on AI topics. | sigecom.org |
| SC4AI (Social Choice for AI) | Workshop series on how social choice can help with AI alignment. | SC4AI |
| Principled Agents | Research nonprofit working on designing safe goals for AGI, using mechanism design, social choice, and game theory. Grew out of the Mechanism Design for AI Safety (MDAIS) reading group, and runs a seminar-style discussion group. | principledagents.org / Discussion group |
| CHAI (Center for Human-Compatible AI) | Stuart Russell's center at UC Berkeley. Frames alignment in decision-theoretic and principal-agent terms; natural fit for economists. | humancompatible.ai |
| Alignment Forum | Central online hub for alignment research discussion and debate. | alignmentforum.org |
| BATES (Bay Area Tech Economics Seminars) | Seminar series at Stanford bringing together economists from academia and industry working on the technology economy, including AI. Free and open. | Stanford Data Science |
| CEPR AI Research Policy Network | Led by Anton Korinek (UVA/Brookings). Webinar series and annual conference. Broad scope, but covers mechanism design, alignment, and compute governance alongside other AI economics topics. | CEPR AI Network |
| GovAI researcher network | The Centre for the Governance of AI at Oxford maintains an active researcher network and runs workshops relevant to compute governance and AI cooperation. | governance.ai |
| Program | Description | Link |
|---|---|---|
| EEML (Eastern European Machine Learning Summer School) | Intensive exposure to cutting-edge ML research with strong networking opportunities. | eeml.eu |
| M2L (Mediterranean Machine Learning Summer School) | Similar format with focus on deep learning and applications. | m2lschool.org |
| Extensive list of ML summer schools | Community-maintained directory. | awesome-mlss |
For economists interested in the intersection of economic theory and AI safety, particularly mechanism design, social choice, and principal-agent problems applied to alignment.
| Program | Description | Link |
|---|---|---|
| BlueDot Impact | Online course providing foundational knowledge in AI safety principles. Good starting point. | bluedot.org |
| MATS (ML Alignment & Theory Scholars) | Competitive fellowship pairing researchers with mentors at leading alignment labs. | matsprogram.org |
| PrincInt (Principles of Intelligence) | Summer fellowship for researchers applying insights from social science and biology to AI safety. Formerly PIBBSS. | princint.ai |
| PIVOTAL | Fellowship program for AI safety research. | pivotal-research.org |
| Constellation | Astra Fellowship supporting independent AI safety research. | constellation.org |
| LASR (London AI Safety Research Labs) | 13-week research program in London. Small teams produce an AI safety paper with expert supervision. | lasrlabs.org |
| Anthropic Fellowship | Fellowship at Anthropic for safety-focused research. | anthropic.com |
| ARENA (AI Research Experience for Academics and Newcomers) | Structured program for researchers transitioning into AI. | arena.education |
| GovAI Summer Fellowship | Summer fellowship at the Centre for the Governance of AI, Oxford. Relevant for economists working on compute governance and mechanism design for AI regulation. | governance.ai |
| AI Safety Camp | Collaborative research program where small teams work on focused AI safety projects over several months. Multiple cohorts per year. | aisafety.camp |
| SPAR (Supervised Program for Alignment Research) | Semester-long, virtual, part-time mentored alignment research, run under the Kairos AI Project. Lower commitment than MATS or PrincInt. | sparai.org |
| CAIRF (Cooperative AI Research Fellowship) | 3-month in-person fellowship in Cape Town. Tracks in multi-agent safety, cooperative AI, and gradual disempowerment. | cai-research-fellowship.com |
Found a dead link, a new fellowship, or a resource that should be here? Please open an issue or submit a pull request. See CONTRIBUTING.md for guidelines.