Research
The AI & Society Lab conducts research about the societal impacts of AI and how various AI governance procedures and technical mechanisms can help to bring about better societal outcomes. This includes work along a variety of ongoing and overlapping societal themes. Work by the lab also includes developing interpretability techniques and using them to further science.
Highlighted
Auditing GPT's Content Moderation Guardrails: Can ChatGPT Write Your Favorite TV Show?
The 2024 ACM Conference on Fairness Accountability and Transparency
·
2024
The (Im)possibility of fairness
Communications of the ACM
·
2021
Fairness and Abstraction in Sociotechnical Systems
Proceedings of the Conference on Fairness, Accountability, and Transparency
·
2019
Certifying and Removing Disparate Impact
Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining
·
2015
Selected in-progress works
Government AI Files
The U.S. federal government releases information about AI systems through a variety of (imperfect) transparency filings. In this project, our goals are to make it easy to search across filings, reveal how the government uses AI, and suggest improved AI transparency mechanisms.
AI Watchman
Chatbots rely on content moderation to keep undesireable content, like violent or sexual content, from being generated. But such filters can also block the generation of other information. In this project, we longitudinally track what societal topics are refused by OpenAI’s GPT series and DeepSeek’s chatbot.
All
2026
Fast algorithms to improve fair information access in networks
Forthcoming, PLOS Complex Systems
·
2026
The OMB Artificial Intelligence Memoranda
Forthcoming, Berkeley Law and Technology Journal
·
2026
2025
Identity-related Speech Suppression in Generative AI Content Moderation
Proceedings of the 5th ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization
·
2025
Feature Responsiveness Scores: Model-Agnostic Explanations for Recourse
International Conference on Learning Representations
·
2025
2024
Auditing GPT's Content Moderation Guardrails: Can ChatGPT Write Your Favorite TV Show?
The 2024 ACM Conference on Fairness Accountability and Transparency
·
2024
2023
Reducing Access Disparities in Networks using Edge Augmentation
2023 ACM Conference on Fairness Accountability and Transparency
·
2023
Measuring and mitigating voting access disparities: a study of race and polling locations in Florida and North Carolina
2023 ACM Conference on Fairness Accountability and Transparency
·
2023
Energy and Carbon Considerations of Fine-Tuning BERT
Findings of the Association for Computational Linguistics: EMNLP 2023
·
2023
2022
Active meta-learning for predicting and selecting perovskite crystallization experiments
The Journal of Chemical Physics
·
2022
Models for understanding and quantifying feedback in societal systems
2022 ACM Conference on Fairness Accountability and Transparency
·
2022
2021
The (Im)possibility of fairness
Communications of the ACM
·
2021
Shapley Residuals: Quantifying the limits of the Shapley value for explanations
Neural Information Processing Systems (NeurIPS)
·
2021
2020
Fairness warnings and fair-MAML: learning fairly with minimal data
Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency
·
2020
Problems with Shapley-value-based explanations as feature importance measures
Proceedings of the 37th International Conference on Machine Learning
·
2020
2019
Gaps in Information Access in Social Networks
The World Wide Web Conference
·
2019
Fairness in representation: quantifying stereotyping as a representational harm
Proceedings of the 2019 SIAM International Conference on Data Mining
·
2019
Assessing the Local Interpretability of Machine Learning Models
NeurIPS Workshop on Human-Centric Machine Learning
·
2019
Energy Usage Reports: Environmental awareness as part of algorithmic accountability
NeurIPS Workshop on Tackling Climate Change with Machine Learning
·
2019
Fairness and Abstraction in Sociotechnical Systems
Proceedings of the Conference on Fairness, Accountability, and Transparency
·
2019
A comparative study of fairness-enhancing interventions in machine learning
Proceedings of the Conference on Fairness, Accountability, and Transparency
·
2019
Automated Congressional Redistricting
ACM Journal of Experimental Algorithmics
·
2019
Disentangling Influence: Using disentangled representations to audit model predictions
Advances in Neural Information Processing Systems
·
2019
2018
Decision Making with Limited Feedback: Error bounds for Recidivism Prediction and Predictive Policing
Proceedings of Algorithmic Learning Theory
·
2018
Interpretable Active Learning
Proceedings of the 1st Conference on Fairness, Accountability and Transparency
·
2018
Runaway Feedback Loops in Predictive Policing
Proceedings of the 1st Conference on Fairness, Accountability and Transparency
·
2018
Auditing black-box models for indirect influence
Knowledge and Information Systems
·
2018
2016
Principles for accountable algorithms and a social impact statement for algorithms
Dagstuhl working group write-up
·
2016
Hiring by Algorithm: Predicting and Preventing Disparate Impact
Presented at the Yale Law School Information Society Project conference Unlocking the Black Box: The Promise and Limits of Algorithmic Accountability in the Professions
·
2016
Convex Hull for Probabilistic Points
2016 29th SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI)
·
2016
2015
Certifying and Removing Disparate Impact
Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining
·
2015
A sensor-based framework for kinetic data compression
Computational Geometry
·
2015
2013
Permissions based on wireless network data
US patent 20130244684 A1
·
2013
Position indication controls for device locations
US patent 20130131973 A1 (also WO 2013078125 A1)
·
2013
2011
2010
Geometric Algorithms for Objects in Motion
Ph.D. thesis from University of Maryland, College Park. Dissertation committee: Prof. David Mount (chair), Prof. William Gasarch, Prof. Samir Khuller, Prof. Steven Selden, Prof. Amitabh Varshney.
·
2010
Spatio-temporal Range Searching over Compressed Kinetic Sensor Data
Lecture Notes in Computer Science
·
2010
Approximation algorithm for the kinetic robust K-center problem
Computational Geometry
·
2010
2009
Compressing Kinetic Data from Sensor Networks
Lecture Notes in Computer Science
·
2009
2008
Enabling teachers to explore grade patterns to identify individual needs and promote fairer student assessment
Computers & Education
·
2008