Detecting and Mitigating Bias by Treating Fairness as a Symmetry Operation
Quick Answer
This study introduces a framework treating bias in machine learning as a symmetry operation, achieving over 90% reduction in violations with a 5% accuracy cost.
Quick Take
This study introduces a framework treating bias in machine learning as a symmetry operation, achieving over 90% reduction in violations with a 5% accuracy cost. It operates without causal graph knowledge and is applicable to any sensitive attribute defined as a bit-flip, making it ideal for high-stakes socio-economic contexts.
Key Points
- Bias is defined as a symmetry breaking operation in classifiers.
- The framework uses loss-based regularization to restore symmetry.
- Evaluated on four synthetic datasets with varying noise and bias.
- Achieves over 90% violation reduction with a 5% accuracy trade-off.
- No causal graph knowledge is required, ensuring computational efficiency.
Article Excerpt
From source RSS / original summaryarXiv:2606. 06514v1 Announce Type: new Abstract: Machine learning systems deployed in high stakes socioeconomic settings routinely display bias. We formalize bias as a symmetry breaking operation: a classifier is fair if its outputs remain invariant under the counterfactual operation of switching a sensitive attribute, with merit features held fixed.
We implement loss based regularization as a symmetry restoring mechanism and evaluate the framework on four synthetic datasets with varying levels of noise, correlation, and bias. The framework achieves upwards of 90\% violation reduction, with accuracy costs around 5\%.
This framework does not require causal graph knowledge, is computationally lightweight, and generalizes to any sensitive attribute definable as a bit-flip, making it suitable for contexts where local sources of discrimination remain absent from mainstream benchmarks.
Reader Mode unavailable (could not extract clean content).
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →The Sim-to-Real Gap of Foundation Model Agents: A Unified MDP Perspective
This paper addresses the sim-to-real gap for foundation model agents by framing it within a Markov Decision Process (MDP) structure. It advocates for established solutions like domain randomization to enhance agent robustness, aiming to create standardized benchmarks for reliable real-world applications.