AI and the Complexity of Pain

Abstract Pain is a complex, multidimensional phenomenon. Pain research documents cases where our disregard of diverse pain experiences leads to epistemic injustices against those who suffer from pain. Automated pain detection technologies use limited behavioural and physiological indicators and are trained on insufficiently diverse datasets. These technologies can potentially exacerbate existing epistemic harms against pain sufferers, particularly marginalised groups. I argue that we must carefully consider the diverse presentations of pain and their neurological, biological, psychological, and social dimensions when implementing AI-based pain assessment tools. To prevent deepening epistemic injustices in pain care, we must develop AI systems by collecting diverse datasets, implementing explainable algorithms, maintaining transparency in data and algorithms, and incorporating flexibility for ongoing improvements. Preserving pain’s complexity when developing these technologies helps prevent further marginalising those who experience pain, particularly vulnerable populations.