![[2602.13286] Explanatory Interactive Machine Learning for Bias Mitigation in Visual Gender Classification](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.13286] Explanatory Interactive Machine Learning for Bias Mitigation in Visual Gender Classification
Summary
This article explores Explanatory Interactive Machine Learning (XIL) as a method to mitigate bias in visual gender classification, demonstrating its effectiveness in improving model fairness and transparency.
Why It Matters
Bias in machine learning, particularly in gender classification, poses ethical concerns and can lead to unfair outcomes. This study highlights innovative approaches to reduce bias through interactive learning, contributing to the development of fairer AI systems.
Key Takeaways
- Explanatory Interactive Learning (XIL) allows user feedback to guide model training.
- The study evaluates two XIL strategies, CAIPI and RRR, and a hybrid approach for bias mitigation.
- Results indicate that CAIPI effectively reduces bias in gender classification while maintaining or improving accuracy.
- Increased transparency from XIL methods can lead to better model fairness.
- The findings support the potential of XIL in addressing ethical concerns in AI.
Computer Science > Computer Vision and Pattern Recognition arXiv:2602.13286 (cs) [Submitted on 7 Feb 2026] Title:Explanatory Interactive Machine Learning for Bias Mitigation in Visual Gender Classification Authors:Nathanya Satriani, Djordje Slijepčević, Markus Schedl, Matthias Zeppelzauer View a PDF of the paper titled Explanatory Interactive Machine Learning for Bias Mitigation in Visual Gender Classification, by Nathanya Satriani and 3 other authors View PDF HTML (experimental) Abstract:Explanatory interactive learning (XIL) enables users to guide model training in machine learning (ML) by providing feedback on the model's explanations, thereby helping it to focus on features that are relevant to the prediction from the user's perspective. In this study, we explore the capability of this learning paradigm to mitigate bias and spurious correlations in visual classifiers, specifically in scenarios prone to data bias, such as gender classification. We investigate two methodologically different state-of-the-art XIL strategies, i.e., CAIPI and Right for the Right Reasons (RRR), as well as a novel hybrid approach that combines both strategies. The results are evaluated quantitatively by comparing segmentation masks with explanations generated using Gradient-weighted Class Activation Mapping (GradCAM) and Bounded Logit Attention (BLA). Experimental results demonstrate the effectiveness of these methods in (i) guiding ML models to focus on relevant image features, particularly w...
