Umeå universitets logga

In recent years, research on explainable AI (henceforth, XAI) has intensified, responding to the societal challenge that many algorithmic approaches (such as machine learning or autonomous intelligent systems) are rapidly increasing in complexity, making it difficult for citizens to understand their assistance and to accept the decisions they suggest. There now exists a large body of approaches pushing forward many ideas of how algorithms should be made explainable or even be able to explain their own output. However, the users’ perspective is rarely taken into account seriously. Obviously, not only research dimensions but also putting into operation a true involvement of the users are missing.

In this meeting, we therefore gathered scholars from different disciplines to account for the question of how explanation generation can be tailored to the users and their way of understanding. Starting this 4-days meeting, we proposed that, rather than being ‘delivered’ by the explainer, explanations become tailored when they emerge at the interface between the explainer (or explaining system) on the one hand and explainee (the receiver of an explanation) on the other hand. Both are active participants in shaping explanations during a social interaction. We thus considered social interaction to be the key to the involvement of the users.