Semantic match: Debugging feature attribution methods in XAI for healthcare
Giovanni Cinà* (Amsterdam University Medical Center), Tabea E. Röber (University of Amsterdam), Rob Goedhart (University of Amsterdam), Ş. İlker Birbil (University of Amsterdam)
Abstract: The recent spike in certified Artificial Intelligence tools for healthcare has renewed the debate around adoption of this technology. One thread of such debate concerns Explainable AI and its promise to render AI devices more transparent and trustworthy. A few voices active in the medical AI space have expressed concerns on the reliability of Explainable AI techniques and especially feature attribution methods, questioning their use and inclusion in guidelines and standards. We characterize the problem as a lack of semantic match between explanations and human understanding. To understand when feature importance can be used reliably, we introduce a distinction between feature importance of low- and high-level features. We argue that for data types where low-level features come endowed with a clear semantics, such as tabular data like Electronic Health Records, semantic match can be obtained, and thus feature attribution methods can still be employed in a meaningful and useful way. For high-level features, we sketch a procedure to test whether semantic match has been achieved.