Position: We Fool Ourselves Thinking the ‘X’ in XAI is Useful

Rolf Hvidtfeldt, Mohammad Naser Sabet Jahromi, Àlex Pujol Vidal, Thomas Gammeltoft-Hansen, Thomas Ploug, Jeppe Agger Nielsen, Stine Nørgaard Christensen, Thomas B. Moeslund

Keywords: XAI, Interdisciplinarity, Epistemic parity, Epistemic power dynamics

TL;DR: The technical paradigms in XAI fail not only in execution but in their fundamental conception.

Abstract: This position paper argues that the prevailing XAI paradigm suffers not from technical limitations but from a profound philosophical misconception: the assumption that explanation is primarily about transparency. We argue that the development of useful explanation is not fundamentally a question of information transfer, but of epistemic parity. Through a case study of harbor monitoring, we demonstrate that although stakeholders are without doubt conceptually misaligned—misalignment primarily results from clashes between domain-specific assumptions and authority claims. We call for replacing transparency-focused XAI with a new paradigm of Domain Authority Negotiation (DAN) that explicitly acknowledges explanation as a contestation of epistemic power rather than a mechanism of revelation.

Submission Number: 396