Photo via iStock/gorodenkoff

Voices & Opinion

AI Could Make Data Collection Less Biased, More Fair

Too much data now ignores under-resourced people

I recently was a panelist at the 2024 Consumer Electronics Show in Las Vegas on the topic of Harnessing the Power of AI Ethically. The American Psychological Association–sponsored discussion came amid the rapid market penetration of ChatGPT. While the two other panelists raised many important concerns about the limitations and misuse of AI, and the impact of what is generated from it, I argued that not using AI was equally an ethical concern.

Given BU’s commitment to addressing issues of diversity, equity, inclusion, and justice across its campuses, we must understand how technology, including use of AI, can help achieve the objective of a more inclusive world. Mobile technologies, particularly applications centered around the smartphone, create such an opportunity, notably for disorders that are heterogeneous in their clinical expression, such as those faced by psychologists. 

Right now, much of the data that health-related research relies on to feed into AI systems comes from those who are highly resourced. This includes the United States, a high-income country in which low-resourced populations are not represented. Globally, it includes low- to middle-income countries (LMIC), keeping in mind that there are those who are highly resourced even in these LMIC. Smartphones are the most market-penetrating technology in the world and have multiple sensors that can serve as digital data collection tools. Essentially, it means many people have a computer in the palm of their hand, and AI can be used to garner insights from any phone’s wide range of digital data streams. 

I’m part of a BU team that’s collaborating with the Davos Alzheimer’s Collaborative to leverage the deep penetration of smartphones [into human lives]. We are collecting digital data through smartphone applications that can be deployed anywhere by anyone—including low-income populations. Importantly, this protocol can be done by individuals at their own place of residence and doesn’t require skilled training to administer. 

Here in the United States, with its relatively robust healthcare system, there is concern about highly trained clinicians inappropriately being replaced by AI. The concern has merit. But consider the clinical expertise available within the United States that is not available to others worldwide. Any digital, AI-driven solution that brings clinical services to regions where there are none available, here and globally, is another important ethical consideration.

Nothing can happen in the use of AI without protecting privacy and confidentiality. This means creating open-source, automated de-identification processing tools that can be made broadly available for anyone to use. Further, the data that will be collected across a broad population will be complex and variable, and current research methods typically used to harmonize disparate datasets aren’t sufficient to fully leverage these multidimensional data resources. Finally, it will not be possible to put all the data into a single data platform. Interoperability between platforms needs to be a priority in order to pull all the data needed to make comprehensive representativeness a reality.

The White House’s Blueprint for an AI Bill of Rights highlights the need for individuals to know when AI is being used and to have options for opting out or addressing problems of misuse. For someone to determine what data they do and do not want to share, it will be important to know: What is the value proposition? Each person needs to decide what is worth giving up in order to get back something of value. A non-health example is a person with a poor sense of direction who chooses to use GPS. Despite knowing that using this technology allows unknown actors to track you, the decision to share this information is considered an appropriate cost in exchange for being able to get somewhere.

Digital is the new blood, and AI is the analytic engine that makes it so. It is well accepted that from a single tube of raw blood there is the potential to measure many markers of disease. Such is the case with digital. Digital, combined with AI, has an opportunity to be truly transformative. 

When encouraging innovation, it is often said there is a need to think out of the box. But what happens if it’s not a box? Relying on what is known, and what has been done to determine what comes next, limits the true promise of AI. The point of technology is not simply to do a better version of what is already being done. The world of psychology, for instance, is challenging, and the power of technology could be used to solve problems that have yet to be solved. 

Current practice is to fit science into known methods. To do anything that will be truly transformative—that is wholly paradigm-shifting—means harnessing the power of AI and developing new methods to fit the science. Only then will solutions for the most intractable, convoluted problems emerge that will serve all, instead of some. 

Rhoda Au is a professor of anatomy and neurobiology at the Chobanian & Avedisian School of Medicine. She can be reached at rhodaau@bu.edu. 

“POV” is an opinion page that provides timely commentaries from students, faculty, and staff on a variety of issues: on-campus, local, state, national, or international. Anyone interested in submitting a piece, which should be about 700 words long, should contact John O’Rourke at orourkej@bu.edu. BU Today reserves the right to reject or edit submissions. The views expressed are solely those of the author and are not intended to represent the views of Boston University.