In Explainable Artificial Intelligence (XAI), local explanation is a popular approach contrasting with global explanation. However explanations by existing local explanation methods always partly agree and partly disagree. Hence individual XAI users have to decide which part of a given explanation to accept and which part to reject. In this paper, assuming that such an acceptability behaviour of a XAI user can be observed over different time points (at each time point, the user applies multiple local explanation methods for the same explanation task), we present an argumentation-based approach for computing and aggregating the user’s revealed preferences over relevant parts of explanations. It is argued that by dealing with the user’s revealed preferences at a level of detail justified by the user’s observed acceptability behaviour, the contribution paves the way to a principled and user-centric framework for resolving conflicts among XAI local methods.