Bias and hallucinations can be unfortunate byproducts of artificial intelligence tools, but one UAB faculty member is turning these challenges into opportunities by leveraging AI to detect implicit and explicit bias in research studies.
Throughout his more than 20-year career, Mohanraj Thirumalai, Ph.D., has dedicated himself to improving the lives of people with disabilities. One of his latest efforts focuses on identifying bias in research studies that could exclude this population, either intentionally or inadvertently. To achieve this, he uses Azure AI — one of the latest AI tools approved for UAB use — to demonstrate how AI can serve as a powerful ally when paired with human insight.
Thirumalai is an associate professor and director of Health Technology in the Division of General Internal Medicine and Population Sciences at the Heersink School of Medicine. He is also the director of UAB’s Rehabilitation Engineering Research Center on Recreation, Sport, and Exercise Technologies for People with Disabilities.
Using Azure AI, Thirumalai analyzes vast amounts of data from the federal clinical trials database, sifting through over half a million records of human studies. The AI-powered analysis identifies research opportunities that could benefit individuals with disabilities while flagging studies that might unintentionally exclude them. For instance, a study requiring participants to stand for extended periods may inadvertently exclude those with mobility impairments, even if standing isn’t essential for the study protocol.
The findings are presented on an interactive dashboard designed to inform researchers, policymakers, and other stakeholders. Thirumalai’s long-term vision includes using this data to influence policy changes and ensuring protections for people with disabilities akin to those in place for other underserved populations. Additionally, the data will help create targeted interventions for researchers to design more inclusive studies.
Azure AI plays a crucial role in streamlining this analysis.
“Given that there are more than half a million records, I can’t just copy and paste into a ChatGPT window,” Thirumalai said. “But I can connect to the OpenAI instance provided by Azure.”
He emphasized the user-friendly nature of Azure AI, noting that researchers only need an API key from UAB IT to begin.
Thirumalai encourages faculty and researchers at UAB to explore the possibilities AI offers across various fields. In his line of research, he envisions AI not only improving research inclusivity but also transforming participant adherence to studies, scaling the number of participants supported by study staff, enhancing patient education, and more.
“The true power of generative AI technologies like ChatGPT lies in their ability to comprehend and generate responses in natural language,” he said. “With vast amounts of data buried within unstructured text across research studies, now is the time to unlock this embedded information to enhance our interventions, refine our processes, and drive innovation.”
As AI tools like UAB-approved Azure expand their capabilities, Thirumalai invites researchers to collaborate on incorporating AI into their studies.
“Whether you’re exploring new interventions, seeking ways to improve participant engagement, or aiming to eliminate bias, AI can be a transformative ally,” he said. "Together, we can leverage AI's potential to enhance research outcomes and create innovative solutions for our challenges."
He is particularly interested in partnerships that explore how AI can be integrated into studies involving underserved populations or complex datasets, bringing new dimensions to medical and social research.