Desirability vs. feasibility
a research through design inquiry of explainable AI
Lorenzo Corti (TU Delft - Electrical Engineering, Mathematics and Computer Science)
Jie Yang (TU Delft - Electrical Engineering, Mathematics and Computer Science)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
Explainability is a key property towards the trustworthiness of AI systems, and scholars have approached the problem from both system-and human-centered angles. However, thus far, these two camps have been researching explainability in isolation without fully accounting for the socio-technical nature of AI systems and the bidirectional effects that emerge when interacting with them. In this chapter, we adopt Research through Design (RtD) -a systematic, reflexive, and documented research approach rooted in design practice -to question our understanding of explainability and how it should be researched. We revisit the explainable AI literature to surface challenges and outline opportunities around bidirectionality, the role of human knowledge, and the availability and temporal dynamics of explanations. We exemplify challenges and opportunities through a speculative exercise grounded in real-world use cases of explainable AI systems. Aligned with RtD, we hope this work helps researchers in explainable AI whom we invite to engage in interdisciplinary and well-documented research processes.