An MDP Approach for Explanation Generation.
Francisco Elizalde, Luis Enrique Sucar, Alberto Reyes, Pablo deBuen · 2007
In order to assist a power plant operator to face un-usual situations, we have developed an intelligent as-sistant that explains the suggested commands generated by an MDP-based planning system. This assistant pro-vides the trainee a better understanding of the recom-mended actions to later generalize them to similar situ-ations. In a first stage, built-in explanations are prede-fined by a domain expert and encapsulated within ex-planation units. When the operator takes an incorrect action, an explanation is automatically generated. A controlled user study in this stage showed that expla-nations have a positive impact on learning. In a second stage, we are developing an automatic explanation gen-eration mechanism based on a factored representation of the decision model used by the planning system. As part of this stage, we describe an algorithm to select a relevant variable, which is a key component of the ex-planations defined by the expert.