Can AI Guide Us Toward a Healthier Future? Currently, It's Still Too Forceful

Artificial intelligence tools, such as ChatGPT, show promise in motivating individuals to adopt healthier lifestyles, yet current implementations often fall short of optimal practice, according to new research from Flinders University. As chronic diseases continue to rise globally, straining healthcare systems, innovative and accessible tools are desperately needed to support health behavior change.

Lead researcher Dr. Candice Oster from Flinders' Caring Futures Institute highlights the potential of AI chatbots as cost-effective options for encouraging lifestyle modifications linked to chronic disease prevention. Nevertheless, evidence of their effectiveness remains limited.

In the study, researchers simulated patient interactions to assess GPT-4's capacity to perform motivational interviewing, a proven counseling technique that helps individuals uncover their own motivation for change. The conversations were evaluated with standard tools for measuring motivational interviewing skills and the results will be presented at the upcoming Australasian Institute of Digital Health's conference, HIC 2025.

Dr. Oster explains that effective motivational interviewing involves a systematic four-step process: building rapport, understanding the individual's goals, eliciting personal motivation, and collaboratively developing a change plan. The initial findings showed that GPT-4 could successfully demonstrate some aspects of motivational interviewing, such as engaging in complex reflections, affirming the individual, and fostering collaboration. It also managed to avoid confrontational responses.

However, the study also uncovered limitations. Over time, the AI tended to shift toward giving lengthy instructions, lists of suggestions, and sometimes attempted prematurely to end conversations or reacted poorly to negative feedback from the simulated patients. These behaviors resemble a more directive style, which can be counterproductive in motivating behavior change, as individuals often resist feeling controlled.

Dr. Oster emphasizes, "Simply telling people what to do is ineffective unless they genuinely want to change. The core challenge is to make AI systems better at supporting autonomous motivation without sounding pushy."

While the initial results demonstrate that GPT-4 has the foundational skills for motivational interviewing, significant improvements are necessary before these tools can be reliably used in health coaching. Researchers suggest that AI models could be refined to better mimic empathetic, person-centered approaches and avoid overly authoritative interactions.

Overall, the study underscores the potential of AI in healthcare support roles but also highlights the need for cautious development to ensure these tools truly enhance patient engagement and promote sustainable health behavior change.