Providing personalized Explanations: a Conversational Approach
Jieting Luo, Thomas Studer, Mehdi Dastani
TL;DR
The paper tackles the challenge of delivering personalized AI explanations by introducing a conversational, justification-logic-based framework that lets an explainer infer an explainee's background from dialogue and iteratively construct tailored explanations. It formalizes a two-agent, modular justification semantics, including learning from explanations via derived terms and a mechanism for explanation selection that prioritizes understandability. A termination guarantee is established under the condition that an understandable explanation exists and the explainer is aware of it, with the approach demonstrated through a chatbot example. The work provides a rigorous foundation for adaptive explanations in AI and outlines future directions for incorporating acceptance criteria and richer modeling of explainee background and values.
Abstract
The increasing applications of AI systems require personalized explanations for their behaviors to various stakeholders since the stakeholders may have various knowledge and backgrounds. In general, a conversation between explainers and explainees not only allows explainers to obtain the explainees' background, but also allows explainees to better understand the explanations. In this paper, we propose an approach for an explainer to communicate personalized explanations to an explainee through having consecutive conversations with the explainee. We prove that the conversation terminates due to the explainee's justification of the initial claim as long as there exists an explanation for the initial claim that the explainee understands and the explainer is aware of.