Evidence is the Antidote to Fear about Artificial Intelligence

For all the promise that Artificial Intelligence (AI) holds, most Americans think it will make their lives worse and will decrease job opportunities. And this isn’t just handwringing by aging Boomers and Gen Xers: Those in their late teens or twenties—Gen Z—are the most concerned about AI, according to a recent Quinnipiac University poll. This widespread dread comes as we are just beginning to see how AI is reshaping how we learn, live, and work. 

The antidote to this fear is evidence. We must do the challenging work of figuring out where AI actually helps, where it does harm, and who bears the costs, either way. Those are questions that my organization and others must help our country answer. 

That urgency for evidence becomes clearer when you consider the good AI can do. It could transform parts of our broken healthcare system that have failed to serve so many. It could expand access to general and specialized medical care to ensure early diagnosis and high-quality chronic disease management. It could quickly guide clinicians to develop and deliver more individualized prevention and treatment strategies for patients. It may help make health services more available to those in rural and other communities that have a shortage of physicians and hospitals. 

This year, my colleagues at the American Institutes for Research (AIR) collected input from researchers, health systems, technology developers, and community organizations to identify where the evidence gaps are and what questions the field still needs to answer. This input will inform a research agenda that can help policymakers, business leaders, and others make better decisions about how to use AI to solve the biggest challenges facing our citizens and our health systems. 

We are also looking at other areas where there is a dearth of evidence, such as education and workforce development. We are studying how AI can help schools, districts, and states make more informed decisions about technology, staffing, and spending and help aspiring tech professionals find jobs, assess adaptive tutoring tools, and connect learners to AI-driven career pathways.

AIR is not alone in this work. The Stanford Institute for Human-Centered AI publishes an annual AI Index, an independent snapshot of where AI actually stands, tracking everything from technical capabilities and economic impact to public opinion and policy. Yale School of Medicine researchers have also been examining whether AI can reliably help doctors predict which emergency room patients are at “at risk of developing agitation in the emergency department.”

For researchers and social scientists, the overall goal should be to keep up with the disruption instead of simply waiting to document its consequences. We should also be thinking long-term. Children starting school today are the first generation that will never know a world without AI. What we learn about them — and from them — will shape policy for generations. We need to know what happens to these children over time as AI is woven into the fabric of our society and our lives.

That is why AIR launched the AI Century Study, which seeks to track how change accumulates as children grow, adults adapt, and communities respond over time. This 100-year national research effort will follow participants from communities across the United States as they grow up, examining how exposure to and use of AI shape cognition and learning, health and well-being, skills and work trajectories, civic engagement, and long-term opportunity. We will release data on an ongoing basis, providing leaders, policymakers, and the public with vital information to inform their decisions. 

Significant blind spots remain and independent, rigorous evidence must meet the moment. AI is already embedded in criminal justice systems, influencing bail recommendations, risk assessments, and law enforcement tools. AI-powered mental health apps and therapy chatbots are proliferating rapidly, yet we have little evidence on whether they help or harm the vulnerable people using them. And as AI shapes what information Americans see online, its effects on civic life, democratic participation, and the spread of misinformation deserve far more rigorous study than they are currently receiving. Closing these evidence gaps will require meaningful investment in independent AI research, from both the private and public sectors.

The public’s skepticism about AI isn’t a problem to be brushed aside; it’s a demand that should be met with data. We need more than marketing and hype. We need tangible evidence. And we can’t wait: AI is already here and reshaping the institutions Americans depend on most. The question isn’t whether to use it. It’s whether we’ll build the evidence base fast enough to use it well.

Jessica Heppen, a researcher and policy expert, is the President & CEO of the American Institutes for Research.