New article by Rune Nyrup (with Luke Connelly, Tine Rønning Pedersen, Klara Øvlisen, Karl-Emil Kjær Bilstrup, and Marianne Graves Petersen)
Explanatory Pragmatism 2.0: Towards a Framework forContext-Sensitive, Human-Centred Interactive Explanations of AI. FAccT '26: The 2026 ACM Conference on Fairness, Accountability, and Transparency (2026)
Abstract
Explainable AI (XAI) and AI literacy research seek to address similar concerns of informed AI use, though these are often pursued in parallel. In this paper, we revisit and refine Explanatory Pragmatism, a context-sensitive XAI framework that describes explanation as a goal-directed, situated activity. Drawing on insights from Activity Theory, broader HCI trends, and AI literacy research, we refine the framework for analysing complex activities and argue for explanations as interaction rather than information transfer. Through three cases spanning feature attribution, interface design, and classroom-based AI literacy, we demonstrate how the framework can be used to bridge XAI, AI Literacy, and the broader practices of explaining AI. We argue that explanation quality depends not only on technical properties such as faithfulness but also on alignment with participants' goals, tools, and contexts, offering a practical lens for evaluating and re-designing explanation practices.