Beyond data-centric explanations for mental healthcare professionals

Abstract

Mental healthcare is facing a capacity crisis where healthcare professionals (HP) do not find the time to interpret the “data overload” generated by longitudinal monitoring tools (wearables and ESM) and do not trust the black-box nature of AI-implemented tools. We developed a mental healthcare dashboard, designed through three iterative user study cycles with HPs in a real-world setting, to bridge this gap. This dashboard synthesizes multiple data streams into a glanceable, summative overview featuring AI-driven stress risk predictions, context-aware recommendations via a RAG-LLM pipeline, and a source-verification mechanism to enable evidence-based reliability. By providing interpretable and interconnected panels, including an interactive assistant for source-grounded queries, our tool aims to reduce cognitive load and increase appropriate trust in this mental health dashboard. This demo highlights how intelligent interfaces can transform raw data into actionable and clinically informed therapeutic insights.

Ähnliche Arbeiten