Revolutionizing Mental Health Support: An Innovative Affective Mobile Framework for Dynamic, Proactive, and Context-Adaptive Conversational Agents

BUILDING A CONTEXT-ADAPTIVE CHATBOT SYSTEM

Our goal for this proposal is to discuss an exploration of the feasibility of building a chatbot system that integrates affective computing and language model-based chatbots to monitor user affective states. This includes adapting context, allowing users to converse and understand why the system categorized the user’s input as a particular emotional state for shared decision support to engage user in cognitive behavioral therapy (CBT). The system will comprehend the text and recognize and respond to the user’s emotional context based on valence and arousal. It will analyze facial behavior, head gestures, pupil-iris ratio, app usage, and physical activity to detect the user’s affective state, creating an interactive agent that provides a more nuanced and emotionally adaptive conversational experience while preserving their transparency and accessibility.

We could also aim to evaluate the interaction’s ability to help users understand how the system works, make informed decisions, judge when to trust the system and learn how accurate it is. To achieve this goal, we can integrate a response decision matrix with a question-driven framework [Liao et al. 2020] and complement it with a clinical decision support system [Schoonderwoerd et al. 2021]. This integration aims to enhance the transparency of the mood prediction model. Specifically, we could utilize the XAI question bank developed by Liao et al. [Liao et al. 2020] to identify a list of explanation methods supported by XAI algorithms.