Counterfactual Fairness

Follow this topic Ask about this topic

Counterfactual fairness studies whether a model treats someone fairly by comparing its prediction with what it would predict for that same person under a change to a sensitive attribute. Causal approaches use structural causal models and path-specific effects to distinguish unfair pathways from justified ones, while model-agnostic methods compare counterfactual input pairs, regularize representations, or adjust predictions after training. The idea is being applied to classification, recommendations, reinforcement learning, graphs, and language-model decisions, where fairness must be balanced against accuracy and utility. A central challenge is that counterfactual fairness does not always guarantee group fairness, and conclusions can depend on uncertain causal assumptions, limited counterfactual data, and noise in model outputs.

Latest papers

22 papers3 papers in the last four weeks, against 2 the four weeks before. 0.0% of all new papers.

Aligning Language Models Towards Counterfactual Invariance
Oct 16, 2026Basar Demir, Emily Shen, Ksheeraj Sai Vepuri +5Demographic Bias in Language Models Counterfactual Fairness
Meta Superintelligence Labs · University of California, San Diego · Independent Researcher
Causally Fair Generation with Large Language Models
Oct 3, 2026Patrik Okanovic, Torsten Hoefler, Drago PleckoAlgorithmic Fairness Causal Interventions in Language Models
Department of Computer Science, ETH Zurich · Department of Statistics & Data Science, UCLA
Minimax Optimal Regret for Causal Logistic Bandits with Counterfactual Fairness
Oct 1, 2026Junhyuk Huh, Seoungbin Bae, Dabeen LeeMulti-Armed Bandits Minimax Regret
Faculty of Mathematics, University of Cambridge · Department of Industrial & Systems Engineering, KAIST · Department of Mathematical Sciences, Seoul National University
Does Thinking Help Fairness? Reasoning Tokens Resolve Some Biases but Create More
Sep 25, 2026Deng Pan, Joe Germino, Yihong Ma +4Social Bias in Language Models Language Model Bias Evaluation
University of Notre Dame · IBM Research, Dublin
Instability Floors: Separating Bias from Noise in Fairness Audits of Clinical LLM Agents with FairMedAgent
Sep 2, 2026Rohith Reddy Bellibatlu, Manpreet Singh, Deepak Parashar +1Counterfactual Evaluation Algorithmic Fairness
Boston University, Boston, MA 02215 USA · Manipal Institute of Technology, Manipal Academy of Higher Education, Manipal, India · Symbiosis International University
Counterfactual Bias Testing for Application Tracking System
Aug 27, 2026Sai Yashwant, Shruti Bansal, Anurag Dubey +4Algorithmic Fairness Algorithmic Bias
ManpowerGroup Services India Pvt. Ltd. · Indian Institute of Technology, Hyderabad
Fairness Invariants: A Relational Approach to Explaining and Mitigating Fairness Bugs
Aug 26, 2026Ranit Debnath Akash, Ashish Kumar, Gang Tan +1Algorithmic Fairness Counterfactual Fairness
University of Illinois at Chicago, Chicago, USA · Pennsylvania State University, University Park, USA
A Distribution Mapping Approach to Counterfactually Fair Reinforcement Learning
Aug 9, 2026Jianhan Zhang, Jitao Wang, John D. Piette +3Causal Counterfactual Generation Algorithmic Fairness
University of Michigan · LSE
GoT-CD: Graph-of-Thoughts Causal Discovery and the Fragility of Post-hoc Path-Specific Fairness Audits
Aug 3, 2026Nitish Nagesh, Elahe Khatibi, Thomas Dean Hughes +3Algorithmic Auditing Algorithmic Fairness
Department of Computer Science, University of California, Irvine, Irvine, CA, USA · University of Orléans, Orléans, France
Do Counterfactually Fair Image Classifiers Satisfy Group Fairness? -- A Theoretical and Empirical Study
Jul 7, 2026Sangwon Jung, Sumin Yu, Sanghyuk Chun +1Algorithmic Fairness Counterfactual Fairness
Department of Electrical and Computer Engineering, Seoul National University · NAVER AI Lab · ASRI/INMC/IPAI/AIIS, Seoul National University
Counterfactual Methods for Detecting Unfairness in Anti-Money Laundering Algorithms
Jul 6, 2026Lea Multerer, Michele Inchingolo, David Kletz +3Algorithmic Fairness Counterfactual Fairness
IDSIA, Switzerland · UBS Switzerland AG and its affiliates, Germany
AgentFairBench: Do LLM Agents Discriminate When They Act?
Jun 15, 2026Triveni Morla, Rohith Reddy Bellibaltu, Manpreet Singh +1Algorithmic Fairness LLM Agent Evaluation
Florida International University, Miami, FL 33199 USA · Boston University, Boston, MA 02215 USA · Department of Computer Science and Engineering, Indian Institute of Technology Patna, Patna, Bihar, India
Detecting and Mitigating Bias by Treating Fairness as a Symmetry Operation
Jun 2, 2026Nishit SinghDebiased ML Algorithmic Fairness
Birla Institute of Technology and Science, Pilani.
COPF: An Online Framework for Deployment-Stable Counterfactual Fairness in Evolving Graphs
May 30, 2026Sheng'en Li, Dongmian ZouAlgorithmic Fairness Counterfactual Fairness
Zu Chongzhi Center and DIRC, Duke Kunshan University, Kunshan, China.
COFT: Counterfactual-Conformal Decoding for Fair Chain-of-Thought Reasoning in Large Language Models
May 28, 2026Arya Fayyazi, Mehdi Kamal, Massoud PedramSocial Bias in Language Models Counterfactual Fairness
Department of Electrical and Computer Engineering, University of Southern California, Los Angeles, California, USA.
Counterfactually Fair Regression via Optimal Transport
May 27, 2026M. Generali Lince, S. Gaucher, J-J. Vie +1Algorithmic Fairness Counterfactual Fairness
Inria, Soda team · CREST, ENSAE, IP Paris · CMAP, CNRS, Ecole Polytechnique, Institut Polytechnique de Paris +1
Automatic Causal Fairness Analysis with LLM-Generated Reporting
Apr 29, 2026Alessia Berarducci, Eric Rossetto, Alessandro Antonucci +1Algorithmic Fairness Counterfactual Fairness
Istituto Dalle Molle di Studi sull’Intelligenza Artificiale (IDSIA), USI-SUPSI, Lugano, Switzerland
Efficient and Interpretable Transformer for Counterfactual Fairness
Apr 29, 2026Panyi Dong, Zhiyu QuanTransformer Interpretability Transformer
Acturial and Risk Mangament Sciences, University of Illinois Urbana-Champaign, Champaign, Illinois, USA.
CAFP: A Post-Processing Framework for Group Fairness via Counterfactual Model Averaging
Apr 8, 2026Irina Arévalo, Marcos OlivaAlgorithmic Fairness Group Fairness
School of Civil Engineering, Universidad Politecnica de Madrid, Profesor Aranguren 3, Madrid, 28040, Spain. · Bain&Company, Velazquez, Madrid, 28006, Spain.
FairLogue: A Toolkit for Intersectional Fairness Analysis in Clinical Machine Learning Models
Apr 6, 2026Nick Souligne, Vignesh SubbianAlgorithmic Fairness Intersectional Fairness
Department of Biomedical Engineering, University of Arizona, 1127 E. James E. Rogers Way, Tucson, AZ, 85721-0020, United States
Mitigating topology biases in Graph Diffusion via Counterfactual Intervention
Mar 2, 2026Wendi Wang, Jiaxi Yang, Yongkang Du +1Graph Generation Graph Diffusion
Purdue University · Pennsylvania State University
SPARC: Scalable Path-Specific Counterfactual Fairness via Causal Conditional Independence
Dec 6, 2024Bowei Tian, Yexiao He, Ziyao Wang +3Debiased ML Algorithmic Fairness
University of Maryland, College Park, MD 20742, USA · Clemson University, Clemson, SC 29634, USA

Oct 16, 2026J/K move · Enter open · S save

Aligning Language Models Towards Counterfactual Invariance

Basar Demir, Emily Shen, Ksheeraj Sai Vepuri, Sourabh Kulkarni+4

Meta Superintelligence Labs · University of California, San Diego · Independent Researcher

Demographic Bias in Language Models Counterfactual Fairness LLM Alignment

Open paper AskPDF arXiv

Small variations in prompt characteristics can meaningfully alter the semantics, tone, and contextual assumptions of large language model outputs. Such characteristics can involve group variables such as demographics or other sensitive latent attributes. Sensitivity to such input variables is undesirable when demographic information is causally irrelevant to the task, but might still be relevant for generation of semantically coherent responses. This makes the task of enforcing counterfactual invariance while preserving model utility difficult. To study this problem, we first show that standard model alignment recipes (such as reinforcement learning) increase differences in response quality across demographics, while improving overall response quality, ultimately amplifying what is known as counterfactual bias. To address this, we develop a policy gradient technique involving a group-level reward function that operates on matched pairs of counterfactual prompts. For completeness, we build a dataset of controlled counterfactual prompt pairs and design an evaluation that separately measures cases where demographic information is causally relevant to the answer. Finally, we validate our findings on several fairness benchmarks. We demonstrate that our methods reduce counterfactual differences while preserving overall response quality and that the models continue to adapt appropriately when such sensitive information is relevant.