The synthesis group Unpacking Concepts unpacks core conceptualizations and their implications in XAI research. Its approach is to analytically examine publications and other documents frequently cited in the XAI community to identify those conceptual alternatives which direct research in specific ways. The aim of this clarification is for researchers  to develop a greater awareness of the implications, historical burdens and context of scientific concepts. Their first project was an analysis of so-called “needs”, as in “users’ needs”, “explanation needs” (if you want to know more about this, see below). Currently, the group is planning a critical analysis of the notion of “explainability” as it was coined and deployed in the DARPA’s XAI project, particularly in their user studies.

— Example: User needs —

Albeit the notion of need is frequently used, there has been little theoretical reflection on the concept and its particular conceptual implications, historical burdens and context of scientific language. Given the theoretical history of the concept in psychology, economics, and social philosophy, the term “explanation need” is quite contra-intuitive. Needs are usually something that people are subjected to. A need is something that simply comes over you and takes (partly) control of your agency, as in the case of basic needs which are given by nature. Higher-level needs are usually discussed in critical terms, in the sense of exposing how these needs are made in the interests of economic actors. In our understanding, the concept of “explanation need” does not fit into either of those theoretical traditions. What is more, XAI is often strongly linked to the idea of empowering people and their agency. But is a concept which traditionally frames people not as actors but as subjects to external forces really a good ally to strengthen this link?

A01, A02, A04, A05, Ö, B03