Artificial Intelligence for Serious Illness Communication: Proactive Approaches to Mitigating Harm.

Serious illness communication is at the core of palliative care, aligning care with patient preferences and improving patient and family experience. Artificial Intelligence (AI) methods have increasingly been applied to palliative care and provide promising opportunities for measuring and enhancing communication (e.g., capturing speech patterns and delivering feedback). Yet, given known disparities in palliative care and the limitations afforded by our natural communication datasets, this task must be approached with caution. Focusing on the study of communication, we consider assumptions that may be baked into our models (e.g., in data, definitions, measurements, and outcomes) and ways to mitigate potential harm across stages of model development-from setting priorities for AI research and applications in our field, to conducting new data collection efforts which are inclusive and more representative, to incorporating patient-family feedback. Transparency and thoughtfulness in this line of innovation may help us leverage AI to provide more equitable, higher-quality serious illness care (see Figure 1).