Artificial Intelligence (AI) and Scientific Research
%20and%20Scientific%20Research.jpg)
Artificial intelligence (AI) is advancing rapidly, transforming nearly every facet of society, including medicine and scientific inquiry. The research community is actively developing tools, protocols, and methodologies to address both the opportunities and challenges of integrating AI and machine learning into scientific work.
AI has already secured a spot on Britannica’s “History of Technology Timeline,” alongside transformative innovations such as irrigation, sailing, gunpowder, printing, the telephone, and the internet. The scientific research community must—and is—swiftly evaluating the implications of deploying AI in research settings to uphold the principles of the Belmont Report.
“When we discuss AI research, we are primarily referring to work aimed at developing tools that substitute for human decision-making. AI development typically involves gathering and utilizing vast amounts of data to train an algorithm to make decisions or predictions within a specific domain,” explained Elisa A. Hurley, PhD, Executive Director of PRIM&R.
The algorithm, she noted, is tested and validated based on the accuracy of its decisions and predictions. “The objective,” Dr. Hurley stated, “is then to apply the AI model to new, real-world data—such as assigning emergency room beds, assessing suicide risk from social media posts, or monitoring workplace stress via remote sensing technologies, to name just three examples.”
“When we discuss AI research, we are primarily referring to work aimed at developing tools that substitute for human decision-making. AI development typically involves gathering and utilizing vast amounts of data to train an algorithm to make decisions or predictions within a specific domain,” explained Elisa A. Hurley, PhD, Executive Director of PRIM&R.
The algorithm, she noted, is tested and validated based on the accuracy of its decisions and predictions. “The objective,” Dr. Hurley stated, “is then to apply the AI model to new, real-world data—such as assigning emergency room beds, assessing suicide risk from social media posts, or monitoring workplace stress via remote sensing technologies, to name just three examples.”
Data Ethics
.jpg)
When asked about the ethical dimensions of AI in scientific and medical research, Jon Herington, PhD, Assistant Professor of Philosophy and Health Humanities and Bioethics at the University of Rochester, emphasized the significance of “data ethics.”
“The basis for accurate and fair machine learning (ML) algorithms lies in representative and responsible datasets,” Dr. Herington told PRIM&R. “If we want ML to foster a healthier and more equitable future, it is our responsibility as scientists to collect data in ways that respect the diverse abilities, objectives, and histories of the people in our communities.”
“Data ethics goes beyond merely avoiding bias—it involves ensuring that the communities from which we collect data retain some control over the process,” Dr. Herington said. “They need to view the research as legitimate and beneficial.”
“One of the most effective ways to act responsibly is to avoid ‘health equity tourism,’ and instead treat our research projects as long-term partnerships with participants—building their capacity alongside our knowledge,” added Dr. Herington, who led a session titled “The Ethical Imperative and Challenges of Working with Diverse Populations in Digital Research” during a PRIM&R workshop on February 22, 2023.
Blurring of Lines
“In today’s digital era, the line between participating in research and going about daily life is blurred, as the data we generate through routine activities could unknowingly become research data,” Dr. Hurley and PRIM&R’s Director of Public Policy, Sangeeta Panicker, PhD, co-wrote in an article published by The National Council of University Research Administrators.
Researchers across various fields—including biomedical, behavioral, cognitive, educational, and social sciences—have used digital technologies to recruit human participants, implement interventions, analyze data, and share findings, a trend intensified during the current COVID-19 pandemic.
The scale at which information from emerging digital technologies can be gathered and analyzed for research differs greatly from traditional, in-person lab experiments. This creates both an opportunity and critical ethical considerations for the use of AI in research.
Algorithms and models are constantly evolving alongside the personal information people generate through their use of digital technologies. The shifting landscape of how personal data is collected, analyzed, and shared raises serious questions about the adequacy of the current ethical framework for research involving human participants.
“There remains considerable scientific and public confusion about the reality and potential of AI. Yet we know from experience that algorithmic-driven, big-data-informed decision-making has, so far, been notoriously riddled with ethical issues, from justice concerns to informational harms,” said Jonathan Beever, PhD, Associate Professor of Ethics and Digital Culture at the University of Central Florida (UCF) and Director and Co-Founder of the UCF Center for Ethics.
“AI is more likely to worsen rather than resolve these issues. It therefore seems wise to adopt a strong precautionary approach rather than a proactionary one—especially when it comes to personal medical and genetic information,” Dr. Beever noted.
Much AI research falls outside the human subjects research oversight framework for three primary reasons, as outlined in a PRIM&R article, “AI and the ‘downstream’ risks of research” (7/14/22). First, the data involved are often collected, owned, and used by commercial entities, which are largely unregulated. Second, research depends on gathering massive amounts of data from social media, apps, internet browsing histories, wearable devices, and electronic health records—although these are data from and about humans, much of it is either deidentified or already publicly available, and thus largely exempt from IRB review. Third, while there may be risks to individuals whose data are included in the large datasets used to train algorithms—reidentification being the most obvious—those risks are considered low.
Community Concerns About AI Ethics
A significant majority of people have expressed concern about the ethical implications of AI in research. Nearly nine out of ten respondents to a PRIM&R LinkedIn poll on this topic indicated they had some level of concern about “the ethical implications of the use of artificial intelligence in medicine and scientific research.”