Researchers at Weill Cornell Medicine-Qatar (WCM-Q) have published a study that employs large language models and data from wearable smart devices to create personalized recommendations aimed at improving student wellbeing and academic performance. The study, led by Dr. Arfan Ahmed, an assistant professor of research in population health sciences, focused on a group of 12 students, gathering data from academic reports, sleep quality surveys, and wearable Fitbit devices.

The students’ activity levels, including step counts, sleep duration, and stress scores, were monitored over a six-week period. A large language model (LLM) called Llama 3, developed by Meta, was utilized to analyze this data. The research team crafted detailed prompts to guide the LLM in interpreting the data and generating tailored recommendations for each student.

The generated advice was evaluated by three independent raters, including healthcare professionals and academics, who assessed the recommendations based on clarity, actionability, and relevance to the students’ data. Although the analysis indicated that the LLM could effectively integrate diverse data sources to produce contextually relevant recommendations, some suggestions received low or average scores for clarity and actionability. The researchers emphasized the importance of maintaining students’ privacy by collecting all data anonymously and cautioned against overwhelming students with excessive feedback. They also highlighted the risk of “model hallucinations,” where the LLM may produce plausible but incorrect information, underscoring the need for safeguards against harmful recommendations in future research.

Dr. Ahmed noted that the study indicated a strong capacity for large language models to analyze various data types and provide personalized recommendations to enhance students’ wellbeing and academic performance. He expressed optimism about the potential for developing useful tools that could not only support students in achieving their full potential but also alleviate pressures on educators and school administrators.

Dr. Javaid Sheikh, dean of WCM-Q, also commented on the research, stating that it illustrates the significant potential of AI and large language models to elevate academic performance and student wellbeing. He cautioned, however, about the careful implementation of AI approaches to ensure the safety and wellbeing of users. The study, titled “Leveraging LLMs and wearables to provide personalized recommendations for enhancing student wellbeing and academic performance through a proof of concept,” has been published in the leading journal, Scientific Reports.