Enhancing Clinical Decision Support: A Heuristic Evaluation of Explainable AI in Healthcare Dashboards

Explainable Artificial Intelligence (XAI) is crucial for enhancing transparency, interpretability and actionability of AI systems, particularly in healthcare. The SAD XAI Dashboard, a clinical decision support (CDS) tool for sepsis-associated delirium (SAD), assists physicians in understanding AI-driven predictions for SAD. Our study aimed to evaluate the XAI dashboard's compliance with an XAI usability heuristics checklist. Three experts (human factors and health informatics) applied the heuristics checklist to the dashboard. The evaluation identified several usability issues, including unclear sequences of actions, non-standard icons, and inconsistent labelling. Key trust and transparency heuristics were also absent from the dashboard. This evaluation highlighted key usability gaps relevant for developers and implementers to be aware of in XAI dashboard design. Future research should focus on understanding the impact of usability heuristics on end-user adoption, particularly those relating to trust and transparency.