A recent study published in the Journal of the American Medical Informatics Association (JAMIA) explores how large language models (LLMs), specifically generative conversational agents (GAs), respond to different motivational states. Researchers Michelle Bak and Jessie Chin from the University of Illinois Urbana-Champaign evaluated three GAs – ChatGPT, Google Bard, and Llama 2 – and found that these models are able to identify users’ motivation states and provide relevant information when individuals have established goals. However, they are less likely to offer guidance when users are hesitant or ambivalent about changing their behavior.
Bak uses the example of an individual with diabetes who is resistant to adopting a more active lifestyle. She explains that GAs should provide information that helps increase awareness about healthy behaviors, engage users emotionally with the changes, and demonstrate how unhealthy habits can impact those around them. Currently, GAs lack specific information on these processes, which puts individuals in a health disadvantage. On the other hand, for individuals committed to changing their physical activity levels, such as those participating in personal fitness training to manage chronic depression, GAs are able to provide relevant information and support.
Professor Chin believes that the significant gap in LLMs’ ability to respond to certain motivation states calls for further research in health promotion using these models. Bak’s research focuses on developing a digital health solution that utilizes natural language processing and psychological theories to promote preventive health behaviors. She holds a bachelor’s degree in sociology from the University of California Los Angeles. Meanwhile, Chin’s research aims to apply social and behavioral sciences theories to design technologies and interactive experiences that promote health communication and behavior across the lifespan. She heads the Adaptive Cognition and Interaction Design (ACTION) Lab at the University of Illinois and holds degrees in psychology, human factors, and educational psychology with a focus on cognitive science in teaching and learning.
In conclusion, the study sheds light on the capabilities and limitations of LLMs in responding to different motivational states when it comes to promoting health behaviors. While these models show promise in helping individuals with established goals, they struggle to provide guidance to those who are hesitant or ambivalent about behavior change. The researchers’ work highlights the need for further exploration in using LLMs for health promotion and suggests potential avenues for future research in the field. Through their research, Bak and Chin aim to leverage natural language processing, psychological theories, and interactive technologies to empower individuals to make positive changes in their health and well-being.
