Empathy is not a Multiple-Choice Test: Critiquing the Claim that AI can Provide Greater Empathy than Healthcare Providers

Stephanie Larson, PhD HEC-C – Case Western Reserve University

Abstract: Responding to claims that AI medical provider programming demonstrates superior empathy to its human physician counterparts, various authors have recently expressed concern that AI – and Chat GPT specifically – will replace us. While such authors are right to be concerned about the use and impact of AI machines in medical practice, it seems they have grabbed the wrong end of the stick. Rather than fearing that AI has eclipsed the power of human empathy and can now offer a superior patient experience, we should instead focus our concern on the increased reliance on purportedly quantitative measures and p-values, when alternative, human centered interactions and evaluations would suffice. The threat to the future of medical practice comes from humans who uncritically accept that AI is “empathetic,” and that a multiple-choice survey that distills the expression of empathy into a five-point scale can measure the import and comfort that a successful patient–physician relationship can provide.

In this session we argue against the increasingly prevalent claim made by various researchers and the tech industry that AI is “empathetic”. We begin by interrogating claims—from such sources as Jonathan Reisman’s recent NYT Opinion and the often-cited 2023 JAMA study by Ayers et al.--that claim ChatGPT is more empathetic than physicians. ChatGPT may have the capacity to replicate the language and text of what humans would identify as an empathetic response. But such language is not tantamount to the relationship of care that a good physician provides.

Keywords: Artificial Intelligence in the Patient Physician Relationship, AI and Empathy, Patient Experience and AI