A Haddonfield Memorial High School junior joined forces with a lab director to publish a 15-page review in the Journal of Artificial Intelligence on Feb. 28 about the possible benefits of AI in clinical trials.
Meanwhile, the school district released a survey of students, staff and parents last month on the possible impact of AI and what direction the district should take on the issue. It showed that 54% of parents – the largest number of respondents – were in favor of AI integration in the schools, followed by 44% of students and 32% of staff.
Mary Miao wrote the first draft of her review over eight months as she handled regular schoolwork and volunteered at Cooper University Hospital. The paper – “Applications of Artificial Intelligence in Clinical Research: A Narrative Review of Recent Advances and Challenges” – addresses the applicability of AI in clinical trials, from the selection of participants to study follow-up.
Miao’s research involved searching articles about AI over a five-year period. Her review addresses how AI used in medical clinical trials can copy a participants’ health data to increase a study’s sample size, scan social media to check the outcomes of trials and use chatbots to answer participants questions during a trial.
Pengpeng Ma, the lab director at the Environmental Molecular Medicine who worked with Miao on the review, said that both of them had to learn about AI in order to write and research the paper. In the course of their research, they collected about 200 to 300 pieces of AI-related writing.
Part of the reason she wanted to write the review, Miao explained, is to counter fear of AI.
” … AI has such a bad rap right now,” she said. “But I think it has so much more potential to revolutionize medicine specifically and I think it deserves that recognition. Like hey, maybe not all AI is bad.”
Ma noted that while it can sometimes be difficult to recruit study participants, AI could compensate.
“Right now, you can use AI to mimic some patients,” he noted. “Based on the real-world data, you can draw from that. AI can help you. Just like AI can provide more candidates, more participants … That’s one of the very good applications for the clinical research.”
In order to reduce error, Ma and Miao chose from three AI models. While the review focuses more on the positives of the technology in clinical trials, it does address some of the drawbacks, too. Like humans, AI can be prone to bias based on the type of data in which it’s trained.
An example would be if an AI model uses mostly data from one ethnic group when selecting medical trial participants, it could fail to include those from other ethnic backgrounds, Ma and Miao observed. That would mean excluding certain groups from the best medical treatments.
“It’s going to widen the medical gap already present,” Maio explained. “Because if (AI) is only trained on one specific group, then all the medicine treatment is going to be tailored to that specific group. And someone who isn’t from that group … isn’t going to get medicine that works for them, because the AI doesn’t have their information to create the medicines that they need.”
Part of the solution, she believes, is for those who conduct clinical trials to be more educated about the downsides and limitations of AI. Her review also points to data privacy concerns, since AI in clinical trials could mean releasing sensitive and private patient data.
The school district’s survey showed that one of the biggest concerns for teachers, staff and students who responded is that AI could be used to cheat, can be inaccurate and may be relied on too much by students. Some survey comments were more critical of AI.
One said, “Students need to be able to do the work themselves before AI can enhance it.”
Others commented on AI’s “enormous costs” and effect on “water use and electricity.”
Yet others comments were positive.
“Not using AI is like not using the Internet in 2000,” argued one respondent.
“AI is a useful tool at its best when the human is well-informed enough to evaluate its output,” another noted.
While Ma doesn’t use AI for clinical research, he does use ChatGPT, Google’s AI and Claude every day to summarize meetings and other office tasks. Miao uses AI for help with homework and getting math questions explained.
Despite her review, Miao said she would prefer to work with patients in family medicine rather than doing more research. But she would be interested in working with Ma again.
The Haddonfield school district recently published the results of an AI survey that it conducted with students, staff and parents.
