CIS 7000: Machine Learning Ecosystems (Fall 2026)

Instructor	Meena Jagadeesan
Lectures	Monday / Wednesday 10:15am - 11:44am (room TBD)
Office Hours	Weekly, TBD

Course Overview

This graduate-level course covers topics in machine learning ecosystems, building up to the research frontier. While machine learning (ML) models are classically analyzed in isolation, ML models interact with a broader ecosystem of other models, humans, companies, and regulators at deployment. How do these interactions impact performance, safety, and society? How can ML models be designed and evaluated to account for these interactions? We will develop a technical toolkit to study these questions, introducing tools from machine learning foundations, game theory, and dynamical systems. Each unit of the course will be grounded in a specific type of ecosystem, ranging from strategic classification, to content recommendation, to LLM agents, to traffic / autonomous vehicles, to the market of ML model-providers.

The class will combine lecture and discussion (here is a tentative schedule and a sampling of relevant papers). The goal of this class is for students to gain exposure to topics on machine learning ecosystems, to build up the technical toolkit to understand the research frontier of this area, to develop presentation skills, and to learn how to think through questions on the fly. There will be no problem sets, exams, or required readings; instead, the main assignments are a hands-on mini-project and a final project. Grading will be based on the final project (70%; 10% proposal + 10% progress report + 20% final report + 30% final presentation), mini-project (20%), and in-class participation (10%). While there are no official prerequisites for this course, students are expected to have mathematical maturity, and students will engage with a range of different technical ideas over the course of the semester.

Schedule

Tentative schedule — subject to change.

Date	Unit	Lecture
Wed, Aug 26	Introduction.	Introduction.
Mon, Aug 31	Unit 1: Strategic Classification.	Motivating applications (bank lending, government resource allocation) and empirical case studies.
Wed, Sep 2	Unit 1: Strategic Classification.	Background on ML Pipeline and Game Theory.
Mon, Sep 7	Labor Day — no class
Wed, Sep 9	Unit 1: Strategic Classification.	Theoretical framework and results.
Mon, Sep 14	Unit 2: Content Recommendation Platforms.	Background on ML Pipeline.
Wed, Sep 16	Unit 2: Content Recommendation Platforms.	Supply-side effects.
Mon, Sep 21	Unit 2: Content Recommendation Platforms.	LLMs used in search and creation.
Wed, Sep 23	Unit 2: Content Recommendation Platforms.	Dynamical systems background and user preferences.
Mon, Sep 28	Unit 2: Content Recommendation Platforms.	Societal impacts and policy.
Wed, Sep 30	Unit 3: LLM agents.	Background on ML Pipeline.
Mon, Oct 5	Unit 3: LLM agents.	Scaling laws and evaluation.
Wed, Oct 7	Unit 3: LLM agents.	Human-LLM Interactions.
Mon, Oct 12	Unit 3: LLM agents.	Multi-agent systems of LLMs.
Wed, Oct 14	Project proposal workshop.
Mon, Oct 19	Unit 3: LLM agents.	Agentic and multi-agent safety.
Wed, Oct 21	Unit 3: LLM agents.	Societal and Economic Impacts.
Mon, Oct 26	Unit 4: Traffic / Autonomous Vehicles.	Congestion and Braess's Paradox.
Wed, Oct 28	Unit 4: Traffic / Autonomous Vehicles.	Dynamical systems and RL background.
Mon, Nov 2	Unit 4: Traffic / Autonomous Vehicles.	Mixed autonomy traffic.
Wed, Nov 4	Project work day (may be rescheduled).
Mon, Nov 9	Unit 4: Traffic / Autonomous Vehicles.	Modern self-driving car ecosystems
Wed, Nov 11	Unit 5: Market of ML Model-Providers.	Economic Background.
Mon, Nov 16	Unit 5: Market of ML Model-Providers.	Monoculture and Competition
Wed, Nov 18	Unit 5: Market of ML Model-Providers.	Background on Online Learning.
Mon, Nov 23	Unit 5: Market of ML Model-Providers.	Incentives from Learning from User Interactions.
Wed, Nov 25	Friday class schedule — no class
Mon, Nov 30	Wrap up.
Wed, Dec 2	Lecture or project presentations.
Mon, Dec 7	Project presentations.

A sampling of relevant papers

Tentative — subject to change.

(C) marks papers that will be covered in class. (*) marks papers that students can choose for their mini-project, and (S) marks useful surveys. New papers may be added to the list in the coming weeks.

Unit 1: Strategic Classification

(C) Moritz Hardt, Nimrod Megiddo, Christos Papadimitriou, and Mary Wootters. Strategic Classification. ITCS 2016.
(C) Daniel Björkegren, Joshua E. Blumenstock, and Samsun Knight. Manipulation-Proof Machine Learning. arXiv 2020.
(*) Jon Kleinberg and Manish Raghavan. How Do Classifiers Induce Agents to Invest Effort Strategically? EC 2019.
(*) Sagi Levanon and Nir Rosenfeld. Strategic Classification Made Practical. ICML 2021.
(*) Lily Hu, Nicole Immorlica, and Jennifer Wortman Vaughan. The Disparate Effects of Strategic Manipulation. FAccT 2019.
(S) Chara Podimata. Incentive-Aware Machine Learning; Robustness, Fairness, Improvement & Causality. ACM SIGecom Exchanges, Vol. 22, No. 2, 2025.
(S) Nir Rosenfeld. Strategic ML: How to Learn With Data That ‘Behaves’. WSDM 2024 (tutorial).

Unit 2: Content Recommendation Platforms

(C) Meena Jagadeesan, Nikhil Garg, and Jacob Steinhardt. Supply-Side Equilibria in Recommender Systems. NeurIPS 2023.
(C) Sarah Dean and Jamie Morgenstern. Preference Dynamics Under Personalized Recommendations. EC 2022.
(C) Fan Yao, Chuanhao Li, Denis Nekipelov, Hongning Wang, and Haifeng Xu. Human vs. Generative AI in Content Creation Competition: Symbiosis or Conflict? ICML 2024.
(C) Yihang Wu, Jiajun Tang, Jinfei Liu, Haifeng Xu, and Fan Yao. Do AI Overviews Benefit Search Engines? An Ecosystem Perspective. arXiv 2026.
(*) Fan Yao, Chuanhao Li, Denis Nekipelov, Hongning Wang, and Haifeng Xu. How Bad is Top-K Recommendation under Competing Content Creators? ICML 2023.
(*) Negin Golrezaei, MohammadTaghi Hajiaghayi, and Suho Shin. Optimal Contest beyond Convexity. STOC 2026.
(*) Boaz Taitler and Omer Ben-Porat. Braess's Paradox of Generative AI. AAAI 2025.
(S) Sarah Dean, Evan Dong, Meena Jagadeesan, and Liu Leqi. Accounting for AI and Users Shaping One Another: The Role of Mathematical Models. TMLR 2025.

Unit 3: LLM Agents

(C) Jared Kaplan, Sam McCandlish, Tom Henighan, et al. Scaling Laws for Neural Language Models. arXiv 2020.
(C) Jordan Hoffmann, Sebastian Borgeaud, Arthur Mensch, et al. Training Compute-Optimal Large Language Models (Chinchilla). NeurIPS 2022.
(C) Thomas Kwa, Ben West, Joel Becker, et al. (METR). Measuring AI Ability to Complete Long Software Tasks. 2025.
(C) Yilun Du, Shuang Li, Antonio Torralba, Joshua B. Tenenbaum, and Igor Mordatch. Improving Factuality and Reasoning in Language Models through Multiagent Debate. ICML 2024.
(C) Matei Zaharia, Omar Khattab, Lingjiao Chen, et al. The Shift from Models to Compound AI Systems. BAIR Blog, 2024.
(C) LangChain. How and When to Build Multi-Agent Systems. LangChain blog, 2025.
(C) Lilian Weng. Reward Hacking in Reinforcement Learning. Lil'Log (blog), 2024.
(C) Sara Fish, Yannai A. Gonczarowski, and Ran I. Shorrer. Algorithmic Collusion by Large Language Models. arXiv 2024.
(C) Gagan Bansal, Tongshuang Wu, Joyce Zhou, Raymond Fok, Besmira Nushi, Ece Kamar, Marco Tulio Ribeiro, and Daniel S. Weld. Does the Whole Exceed its Parts? The Effect of AI Explanations on Complementary Team Performance. CHI 2021.
(C) Hamsa Bastani and Gerard P. Cachon. The Human-AI Contracting Paradox. SSRN working paper, 2025.
(C) Aaron Chatterji, Tom Cunningham, David J. Deming, Zoë Hitzig, Christopher Ong, Carl Shan, and Kevin Wadman. How People Use ChatGPT. NBER Working Paper, 2025.
(C) David M. Rothschild, Markus Mobius, Jake M. Hofman, Eleanor W. Dillon, Daniel G. Goldstein, Nicole Immorlica, Sonia Jaffe, Brendan Lucier, Aleksandrs Slivkins, and Matthew Vogel. The Agentic Economy. Communications of the ACM (Viewpoint), 2026.
(*) Sayash Kapoor, Benedikt Stroebl, Peter Kirgis, et al. Holistic Agent Leaderboard: The Missing Infrastructure for AI Agent Evaluation. arXiv 2025.
(*) Mert Cemri, Melissa Z. Pan, Shuyi Yang, Lakshya A. Agrawal, et al. Why Do Multi-Agent LLM Systems Fail? arXiv 2025.
(*) Joon Sung Park, Joseph C. O'Brien, Carrie J. Cai, Meredith Ringel Morris, Percy Liang, and Michael S. Bernstein. Generative Agents: Interactive Simulacra of Human Behavior. UIST 2023.
(*) Sumeet Ramesh Motwani, Mikhail Baranchuk, Martin Strohmeier, Vijay Bolina, Philip Torr, Lewis Hammond, and Christian Schroeder de Witt. Secret Collusion among AI Agents: Multi-Agent Deception via Steganography. NeurIPS 2024.
(*) Natalie Shapira, Chris Wendler, Avery Yen, et al. Agents of Chaos. arXiv 2026.
(*) Myra Cheng, Cinoo Lee, Pranav Khadpe, Sunny Yu, Dyllan Han, and Dan Jurafsky. Sycophantic AI Decreases Prosocial Intentions and Promotes Dependence. Science 2026.
(*) Brendan Lucier, Nicole Immorlica, Markus Mobius, Aleksandrs Slivkins, Daniel G. Goldstein, Jake M. Hofman, Sonia Jaffe, and David M. Rothschild. Agentic Markets: Equilibrium Effects of Improving Consumer Search. arXiv 2026.
(S) Lewis Hammond, Alan Chan, Jesse Clifton, et al. Multi-Agent Risks from Advanced AI. 2025.

Unit 4: Traffic / Autonomous Vehicles

(C) David Easley and Jon Kleinberg. Networks, Crowds, and Markets: Reasoning about a Highly Connected World, Chapter 8: Modeling Network Traffic using Game Theory. Cambridge University Press, 2010.
(C) Dorsa Sadigh, Shankar Sastry, Sanjit A. Seshia, and Anca D. Dragan. Planning for Autonomous Cars that Leverage Effects on Human Actions. RSS 2016.
(C) Cathy Wu, Aboudy Kreidieh, Eugene Vinitsky, and Alexandre M. Bayen. Emergent Behaviors in Mixed-Autonomy Traffic. CoRL 2017.
(*) Erdem Bıyık, Daniel A. Lazar, Ramtin Pedarsani, and Dorsa Sadigh. Incentivizing Efficient Equilibria in Traffic Networks with Mixed Autonomy. IEEE Transactions on Control of Network Systems, 2021.
(*) Cathy Wu, Alexandre M. Bayen, and Ankur Mehta. Stabilizing Traffic with Autonomous Vehicles. ICRA 2018.
(*) Daron Acemoglu, Ali Makhdoumi, Azarakhsh Malekian, and Asuman Ozdaglar. Informational Braess' Paradox: The Effect of Information on Traffic Congestion. Operations Research, 2018.

Unit 5: Market of ML Model-Providers

(C) Omer Ben-Porat and Moshe Tennenholtz. Regression Equilibrium. EC 2019.
(C) Jon Kleinberg and Manish Raghavan. Algorithmic Monoculture and Social Welfare. PNAS 2021.
(C) Meena Jagadeesan, Michael I. Jordan, Jacob Steinhardt, and Nika Haghtalab. Improved Bayes Risk Can Yield Reduced Social Welfare Under Competition. NeurIPS 2023.
(C) Aleksandrs Slivkins. Introduction to Multi-Armed Bandits, Chapter 11: Bandits and Agents. Foundations and Trends in Machine Learning, 2019.
(C) Yishay Mansour, Aleksandrs Slivkins, and Zhiwei Steven Wu. Competing Bandits: Learning Under Competition. ITCS 2018.
(*) Jinshuo Dong, Hadi Elzayn, Shahin Jabbari, Michael Kearns, and Zachary Schutzman. Equilibrium Characterization for Data Acquisition Games. IJCAI 2019.
(*) Meena Jagadeesan, Michael I. Jordan, and Jacob Steinhardt. Safety versus Performance: How Multi-Objective Learning Reduces Barriers to Market Entry. PNAS 2025.
(*) Benjamin Laufer, Jon Kleinberg, and Hoda Heidari. Fine-Tuning Games: Bargaining and Adaptation for General-Purpose Models. WWW 2024.
(*) Ander Artola Velasco, Dimitrios Rontogiannis, Stratis Tsirtsis, and Manuel Gomez-Rodriguez. Test-Time Compute Games. arXiv 2026.
(*) Jens Prüfer and Christoph Schottmüller. Competing with Big Data. The Journal of Industrial Economics, 2021.

Assignments

Mini-Project

Each student will pick one of the papers from this list that is marked with a (*), do a mini-project based on the paper, and present their findings to the class. Mini-project presentations will take place at the beginning of class and be spread out throughout the semester. We will have a sign-up sheet to ensure that all students choose different papers.

The mini-project should reflect your engagement with the paper, and be centered around a (small-scale) new result that is not directly present in the paper. It can be:

a simulation that reproduces one of the main results,
a mathematical example that shows the intuition for a theorem or an empirical result, or
a simulation or example that relaxes some of the assumptions in the paper, showing how the results extend or break down.

You'll share your findings from your mini-project with the class. First, you'll give a 8 minute presentation, which will consist of a 4 minute overview of the paper, and 4 minutes on your (small-scale) result. After that, we'll spend about 8 minutes on Q&A about the paper and the project. You should expect in-depth questions about your result and about the setup of the paper. Following your presentation, I'll set up a 10-15 minute meeting with you to give you feedback on your presentation.

Final project

Students will conduct a final project geared towards original research in the area of machine learning ecosystems. Students can work individually or in groups of 2.

Your final project can take a range of different forms, including:

Pick one of the lines of work studied in class. Investigate how to extend the model to incorporate a novel form of interaction or to relax one of the assumptions. Perform an analysis of your model extension either mathematically or empirically.
Pick one of the lines of work studied in class which make simplifying assumptions about how the underlying machine learning model works. Perform an empirical evaluation with a realistic ML model to investigate the gap between the theoretical findings and empirical reality.
Apply the ideas in this class to a machine learning ecosystem that was not covered in class. Look for related work (if there is any). Identify a novel interaction mode between ML models, humans, and companies. Perform a preliminary mathematical or empirical analysis.

Students will submit a 1-page proposal, a 2-page progress report, and a 5 page final report. Students will also give a final presentation at the end of the semester. The length of the presentation slot will be determined by the number of projects. A third of the presentation slot will be devoted to in-depth questions (by me and by the class) about the motivation and technical details of the project; this will give you an opportunity to explain your thought process.