Beyond Feature Overload: Real-Time, Emotion-Aware Explainable AI Paraphrasing for NNESs

Building on Kim et al.’s finding that NNESs use multiple information aids but that these are not equally decisive—and that the combination, not just presence, of aids informs user decision-making—this research explores dynamic adaptation: can the tool sense when a user is overwhelmed, anxious, or confused, and adjust the amount and type of information offered in response? The approach incorporates real-time signals such as facial expressions (via webcam-based emotion detection), keyboard/mouse hesitations, and optional physiological measures like skin conductance to adaptively throttle or elaborate information aids. For example, detecting user anxiety might lead the system to reduce cognitive load by presenting only the most salient suggestion, offering simpler explanations, or pausing to invite clarification. Conversely, if the user appears calm and engaged, the system could offer more detailed comparative explanations or enable exploratory features to support deeper learning and autonomy. This approach differs from prior static, user-driven workflows by integrating affect-aware, co-active assistance that addresses foreign language anxiety and cognitive overload, potentially improving writing outcomes, language acquisition, and user confidence. It also introduces new evaluation dimensions measuring changes in user affect, stress, and confidence, opening avenues for quantitative and qualitative study.

References:

1. Design Opportunities for Explainable AI Paraphrasing Tools: A User Study with Non-native English Speakers. Yewon Kim, Thanh-Long V. Le, Donghwi Kim, Mina Lee, Sung-Ju Lee (2024). Conference on Designing Interactive Systems.