Aburizik to study AI’s ability to detect empathy

Can artificial intelligence (AI) detect and monitor empathy? Arwa Aburizik, MD, MS, clinical associate professor in Hematology, Oncology, and Blood & Marrow Transplantation, has been studying provider wellness and their empathy and attitudes toward the mental health of people with cancer. Previously, she has used self-reported measures and assessment, but now, armed with a pilot grant from the UI’s Iowa Initiative for Artificial Intelligence (IIAI) with AI and machine learning (ML). The IIAI has funded numerous AI/ML multidisciplinary projects over the last 4 years. In collaboration with the IIAI engineers, Aburizik and her research partner, clinical psychologist Martin Kivlighan, PhD, the team will add another to their portfolio.

Aburizik is an internist and psychiatrist and directs the Psychosocial-Oncology Program in the UI’s Holden Comprehensive Cancer Center, which provides holistic care for people with cancer. She and her research team will develop models to use ML/AI in identifying and measuring empathy and attitudes toward mental health among oncologists. Once identified and measured, the knowledge could be used to support efforts to increase and nurture this attribute in providers. “Empathy is both a cause and a result of a positive patient-doctor relationship,” Aburizik said, “and a factor in a good work environment for oncology providers.”

There are many validated measures for empathy. “As empathy varies among clinicians and in the same clinician over time, psychometric assessments are not practical or realistic for continuous evaluation.” But, if changes in empathy could be detected efficiently and continuously via ML/AI, then targeted interventions to help clinicians re-calibrate could be implemented. “More empathy means more positive patient outcomes. This means greater job satisfaction and less turnover of oncology providers. It’s a win for everyone!” Aburizik said.

The research, Aburizik made clear, has been IRB-approved with both patient- and provider-identifying information stripped from the scripts processed by the AI/ML system. The system will also process self-reported data from surveys completed by oncologists related to their knowledge, behavior, and attitudes toward mental health and empathic care, which the Psychosocial-Oncology Program has previously developed. As for future applicability, Aburizik clarifies that any real-world deployment would be an “opt-in” program, with participants providing consent to be assessed and supported. Ethical guidelines would also be in place for how the data is used.

Aburizik’s previous AI/ML project was sponsored by the UI’s Adolescent and Young Adult Cancer Program, whose patients often experience a higher burden of practical and psychological distress. Aburizik and her team examined trends in providers’ ability to be attuned to those unique needs as reflected by orders and referrals post-visit.