Gender Fairness of Machine Learning Algorithms for Pain Detection

2025-06-10

AI Summary

This study systematically investigates gender bias in machine learning and deep learning models for automatic pain recognition from facial expressions. Using the UNBC-McMaster dataset, we benchmark four representative models—linear and RBF-kernel SVMs, CNNs, and Vision Transformers (ViTs)—and evaluate their cross-gender performance via multiple fairness metrics, including equal opportunity difference and predictive equality. Our analysis reveals a significant accuracy–fairness trade-off: while ViT achieves the highest accuracy (78.3%) and excels on certain fairness dimensions, all models exhibit statistically significant gender bias, with an average equal opportunity difference of 12.6%. This confirms a pervasive fairness deficiency in current vision-based pain recognition systems. The work advances medical AI fairness assessment by shifting the paradigm from accuracy-centric evaluation toward a multidimensional, fairness–efficacy co-optimization framework.

Technology Category

Computer Vision: Bias, Fairness & PrivacyMachine Learning: Ethics, Bias, and FairnessPhilosophy and Ethics of AI: Bias, Fairness & Equity

Application Category

Social Networks and Social Media: Fairness and bias in social network and social media analysisUser Modeling, Personalization and Recommendation: Fairness-aware retrieval and rankingEconomics, Online Markets and Human Computation: Fairness and ethical considerations in crowd work and in human-in-the-loop AI systems

Abstract

Automated pain detection through machine learning (ML) and deep learning (DL) algorithms holds significant potential in healthcare, particularly for patients unable to self-report pain levels. However, the accuracy and fairness of these algorithms across different demographic groups (e.g., gender) remain under-researched. This paper investigates the gender fairness of ML and DL models trained on the UNBC-McMaster Shoulder Pain Expression Archive Database, evaluating the performance of various models in detecting pain based solely on the visual modality of participants' facial expressions. We compare traditional ML algorithms, Linear Support Vector Machine (L SVM) and Radial Basis Function SVM (RBF SVM), with DL methods, Convolutional Neural Network (CNN) and Vision Transformer (ViT), using a range of performance and fairness metrics. While ViT achieved the highest accuracy and a selection of fairness metrics, all models exhibited gender-based biases. These findings highlight the persistent trade-off between accuracy and fairness, emphasising the need for fairness-aware techniques to mitigate biases in automated healthcare systems.

Problem

• Investigates gender fairness in ML pain detection models
• Evaluates bias in facial expression-based pain algorithms
• Compares accuracy-fairness trade-offs across ML/DL methods

Innovation

• Evaluates ML and DL models for pain detection
• Compares SVM, CNN, and ViT using fairness metrics
• Highlights fairness-aware techniques to reduce biases