Behavioral Machine Learning Lab

About the lab

Led by Prof. Nir Rosenfeld · Technion CS

Predictive models are deployed among people who have their own goals — applicants, sellers, users, agents. When they adapt to a model, the model changes the world it learned from. We build the theory and tools for learning in these settings: classifiers that account for strategic behavior, predictions that shape the outcomes they forecast, and language models whose incentives align with the people they serve.

GitHub

Nir Rosenfeld

Principal Investigator · Associate Professor, Technion CS

Nir's research lies at the intersection of machine learning and human behavior: modeling and predicting behavior and social dynamics, welfare-centric machine learning, strategic classification and performative prediction, and the implications of deploying predictive models in social contexts. Before joining the Technion he was a postdoc at Harvard SEAS with David Parkes and Yaron Singer, and received his BSc, MA, and PhD from the Hebrew University of Jerusalem, advised by Amir Globerson.

Personal page nirr@cs.technion.ac.il Google Scholar

Publications

2026

Welfare-Optimal Classification with Accuracy Auctions

Bana Sadi*, Eden Saig*, Nir Rosenfeld

ICML 2026 · Spotlight

Behavioral Machine Learning

About the lab

Nir Rosenfeld

Publications

Welfare-Optimal Classification with Accuracy Auctions

Ambiguous Strategic Classification

The Role of Causal Features in Strategic Classification for Robustness and Alignment

Strategic PAC Learnability via Geometric Definability

Welfare as a Guiding Principle for Machine Learning — From Compass, to Lens, to Roadmap

Strategic Classification with Non-Linear Classifiers

Evolutionary Prediction Games

A Market for Accuracy: Classification Under Competition

Learning Classifiers That Induce Markets

Adversaries With Incentives: A Strategic Alternative to Adversarial Robustness

Classification Under Strategic Self-Selection

One-Shot Strategic Classification Under Unknown Costs

Instructed to Bias: Instruction-Tuned Language Models Exhibit Emergent Cognitive Bias

Decongestion by Representation: Learning to Improve Economic Welfare in Marketplaces

Delegated Classification

Causal Strategic Classification: A Tale of Two Shifts

Performative Recommendation: Diversifying Content via Strategic Incentives

Learning to Suggest Breaks: Sustainable Optimization of Long-Term User Engagement

Strategic Classification with Graph Neural Networks

In the Eye of the Beholder: Robust Prediction with Causal User Modeling

Generalized Strategic Classification and the Case of Aligned Incentives

Strategic Representation

Strategic Classification Made Practical

Strategic Classification in the Dark

Lab members

Alumni