Keynote B abstract

Intentional and mechanistic explanation in the context of AI

Marta Halina

University Associate Professor of Philosophy of Cognitive Science

University of Cambridge

18:00–19:30, 15 Dec 2023

In this talk, I consider two approaches to answering the question: When we use AI as an explanatory model, should we require transparency? The first approach rejects the idea that we require transparency in the sense of understanding the inner workings of AI systems. Instead, advocates of this position hold that we should aim for intentional explanations in those cases where AI is replacing human decision-makers. As a second approach, I apply mechanistic accounts of explanation to this question. Such accounts of explanation are popular in philosophy of science but are not typically applied to cases of AI. Doing so reveals that understanding the inner workings of AI is indeed beneficial in many cases where we seek scientific explanations. Finally, I argue that although neither intentional nor mechanistic approaches should be applied across the board, both offer important insights regarding the procedures we should follow for determining the transparency required from AI in the context of scientific explanation and understanding.