A Taxonomy of User-centered Explainable AI Studies

Authors

Lukas-Valentin Herm, University of WürzburgFollow
Jonas Wanner, University of WürzburgFollow
Christian Janiesch, TU Dortmund UniversityFollow

Paper Number

1185

Abstract

The progress in the research field of machine learning is fast-paced and it is most noticeable in terms of prediction performance. However, there seems to be a lack of understanding of the explanatory value for the actual user. As only a user-appropriate implementation realizes effective human-machine cooperation, this must be the goal for any intended intelligent system development. Accordingly, some studies have addressed the problem. However, their aims and methods vary, and a meta synthesis of the results is missing. To address these problems, we have developed a taxonomy of user-centered XAI studies. It allows both the conception and the classification of current user-centered XAI studies. Furthermore, through descriptive analytics and a cluster analysis, we identify patterns and archetypes to better conceptualize the field and support future research.

Comments

Paper Number 1185

Recommended Citation

Herm, Lukas-Valentin; Wanner, Jonas; and Janiesch, Christian, "A Taxonomy of User-centered Explainable AI Studies" (2022). PACIS 2022 Proceedings. 9.
https://aisel.aisnet.org/pacis2022/9