Contextual “Explanation Playlists” Based on Task, Mood, and Cognitive State

Kim et al. (2024) and Kim et al. (2024, Frontiers AI) note the importance of context (e.g., task, user mood, cognitive load) in explanation effectiveness, but most systems offer static or single-format explanations. This idea draws inspiration from music or video playlists: the XAI interface continuously assesses user state through lightweight sensing (e.g., pauses, facial expressions, explicit feedback) and then assembles a tailored sequence of explanation types—visual, textual, analogical, etc.—optimized for the current context. For example, a user under time pressure might get a concise summary, while a reflective user might get a deeper causal trace. This approach would operationalize the “selective” heuristic at a granular level and could be particularly impactful in high-stress or variable environments (e.g., emergency medicine, financial trading). It extends the personalization explored by Mahya & Fürnkranz (2025) from static user preferences to dynamic, context-aware adaptation.

References:

1. Biasing Rule-Based Explanations Towards User Preferences. Parisa Mahya, Johannes Fürnkranz (2025). Information.
2. Human-centered evaluation of explainable AI applications: a systematic review. Jenia Kim, Henry Maathuis, Danielle Sent (2024). Frontiers Artif. Intell..