Semantic match: Debugging feature attribution methods \titlebreak in XAI for healthcare

Giovanni Cina, Tabea E Rober, Rob Goedhard, S Ilker Birbil
Proceedings of the Conference on Health, Inference, and Learning, PMLR 209:182-190, 2023.

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.

Cite this Paper

BibTeX
@InProceedings{pmlr-v209-cina23a, title = {Semantic match: Debugging feature attribution methods \titlebreak in XAI for healthcare}, author = {Cina, Giovanni and Rober, Tabea E and Goedhard, Rob and Birbil, S Ilker}, booktitle = {Proceedings of the Conference on Health, Inference, and Learning}, pages = {182--190}, year = {2023}, editor = {Mortazavi, Bobak J. and Sarker, Tasmie and Beam, Andrew and Ho, Joyce C.}, volume = {209}, series = {Proceedings of Machine Learning Research}, month = {22 Jun--24 Jun}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v209/cina23a/cina23a.pdf}, url = {https://proceedings.mlr.press/v209/cina23a.html}, 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.} }
Endnote
%0 Conference Paper %T Semantic match: Debugging feature attribution methods \titlebreak in XAI for healthcare %A Giovanni Cina %A Tabea E Rober %A Rob Goedhard %A S Ilker Birbil %B Proceedings of the Conference on Health, Inference, and Learning %C Proceedings of Machine Learning Research %D 2023 %E Bobak J. Mortazavi %E Tasmie Sarker %E Andrew Beam %E Joyce C. Ho %F pmlr-v209-cina23a %I PMLR %P 182--190 %U https://proceedings.mlr.press/v209/cina23a.html %V 209 %X 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.
APA
Cina, G., Rober, T.E., Goedhard, R. & Birbil, S.I.. (2023). Semantic match: Debugging feature attribution methods \titlebreak in XAI for healthcare. Proceedings of the Conference on Health, Inference, and Learning, in Proceedings of Machine Learning Research 209:182-190 Available from https://proceedings.mlr.press/v209/cina23a.html.

Related Material