Chloe R. Bennett, Heather Cole-Lewis, Stephanie Farquhar, Naama Haamel, Boris Babenko, Oran Lang, Mat Fleck, Ilana Traynis, Charles Lau, Ivor Horn, Courtney Lyles
Diverse expertise improves AI fairness: The paper argues that bringing together experts from different fields—not just computer scientists—is essential for identifying and reducing bias in AI systems used in healthcare. Interdisciplinary panels can catch problems that single-discipline teams might miss.
Data representation alone isn't enough: While previous research has focused on ensuring representative training data, this paper emphasizes that explainability also matters for equity. Even well-performing models need to be understood through multiple lenses to identify whether they're making decisions for the right reasons.
Historical and contextual knowledge is critical: The framework highlights that AI explanations must be evaluated by experts who understand the social, historical, and structural contexts in which healthcare disparities occur. This helps distinguish between legitimate medical associations and patterns that reflect existing inequities in society.
Bias detection requires collaborative review: The paper proposes that interdisciplinary expert panels can identify potential confounders and gaps in research that might otherwise be overlooked, helping teams understand whether an AI model's explanations are scientifically sound or potentially perpetuating harm.
Better explanations lead to better models: By critically assessing why an AI system makes specific predictions through multiple expert perspectives, organizations can identify concrete opportunities to improve their models and address equity gaps rather than simply deploying systems as-is.
The field of artificial intelligence (AI) is rapidly influencing health and healthcare, but bias and poor performance persists for populations who face widespread structural oppression. Previous work has clearly outlined the need for more rigorous attention to data representativeness and model performance to advance equity and reduce bias. However, there is an opportunity to also improve the explainability of AI by leveraging best practices of social epidemiology and health equity to help us develop hypotheses for associations found. In this paper, we focus on explainable AI (XAI) and describe a framework for interdisciplinary expert panel review to discuss and critically assess AI model explanations from multiple perspectives and identify areas of bias and directions for future research. We emphasize the importance of the interdisciplinary expert panel to produce more accurate, equitable interpretations which are historically and contextually informed. Interdisciplinary panel discussions can help reduce bias, identify potential confounders, and identify opportunities for additional research where there are gaps in the literature. In turn, these insights can suggest opportunities for AI model improvement.
Citation data from Semantic Scholar