While others argue over whether artificial intelligence (AI) will save or doom humanity, Emory is shaping how it serves us. Emory is embracing AI as a powerful tool, investing in its potential and ensuring it benefits everyone. 

Emory research spans nearly every dimension of human inquiry. AI touches them all. Faculty and students are using AI intentionally, teaching it responsibly and advancing it with a clear principle: Technology must follow human purpose — not the other way around. 

Lauren Klein, Winship Distinguished Research Professor at Emory College of Arts and Sciences, says the most promising AI systems have yet to be invented. She is wary of the current wave of for-profit AI software. 

“What we’re being told is AI is coming from a narrow set of people and a handful of corporations that have vastly different incentives than the rest of the world,” she says of the tech firms inserting AI capability into nearly every computer interface and transaction.

Klein is part of a large group of Emory scholars striving to redefine what AI is and how it should work. She and her colleagues want to make sure people, not tech firms and billionaires, determine how AI computing is used in their lives.  “Academic institutions, and Emory researchers in particular, are imagining uses of AI that benefit humanity, increase our understanding of cultures, and are wanted and shaped by communities using them,” she says.

INTENTION, NOT DISRUPTION

Rather than adopting Silicon Valley’s “move fast and break things” ethos, Emory is charting a deliberate path. Faculty approach AI with careful attention to both risks and opportunities, ensuring that new technologies advance learning and human well-being.

Nabile Safdar, chief AI officer for Emory Healthcare, says AI, and how to educate with it, are divisive topics within medicine. Faculty members in his specialty, radiology, have sorted themselves into opposing groups when considering AI access for trainees.  

“One camp asks ‘How will they learn if they’re using AI?’ and the other asks ‘How will they learn if they’re not using AI?’ The task for educators is to find the balance,” Safdar says.

He is sympathetic to both camps. Safdar says there is a real risk of what he calls “deskilling” medical students and practitioners if their education leaves them overly reliant on technology. However, there is equal risk if they’re not taught how to use increasingly widespread tools.

For example, in radiology, he thinks all students should learn the basics of AI imaging so they can use it effectively. AI can help doctors identify information in scans that is not always visible to the eye. Safdar says AI’s role in imaging is to bring new and important information to the attention of clinicians so that they can review and respond to it. 

He co-leads a group advising Emory University and Emory Healthcare on the critical questions posed by AI. He advocates for AI adoption across the university grounded in intention and ethics. Ultimately, people, not tools, shoulder the responsibility for patient care, he says. AI is neither a colleague nor an expert. An AI-enabled medical tool is like a scalpel or an x-ray machine — a tool that a trained clinician can use to help a patient.

Whether practicing or in training, clinicians should understand AI’s inherent limitations, such as its tendency to “hallucinate” and interpret imaging artifacts incorrectly. Also, machine-learning models trained with data from one group of people may not deliver accurate diagnostic information for people with different characteristics. A good AI medical curriculum, Safdar says, starts with safety and ethical training.  

“Too many people are excluded from the advances of modern medicine,” he says. “It doesn’t have to be that way. We can make smarter tools and deliver them more widely. We just have to care enough to try.”

Anant Madabhushi’s pioneering medical imaging work exemplifies the ideal Safdar describes. As executive director for the Emory Empathetic AI for Health Institute, Madabhushi and his research team are building AI algorithms that analyze digitized pathology slides and imaging data to detect cancer, gauge its aggressiveness and help determine the most effective treatment options. The technologies they are creating are not just for top-tier cancer centers, but also for safety-net hospitals and clinics in low-resource communities around the world.

Emory research spans nearly every dimension of human inquiry. AI touches them all.

Personal experience informs Madabhushi’s focus. During the 1990s in India, he lost his beloved aunt to breast cancer. He describes this as an inflection point that continues to inspire his career today. 

“Equity is not a side-benefit. It’s the goal,” Madabhushi explains. “Our systems are trained to work across race, gender and geography.”

BETTER INPUTS, BETTER OUTPUTS

Another hallmark of Emory’s AI approach is inclusion. Klein says one way to improve AI tools is to make sure the people who use them are part of their creation — because inclusive design and construction make tools better. “AI models are trained on data that largely comes from the Internet circa 2010 to now,” she explains. As a result, the tools reflect only a “narrow sliver” of human knowledge and experience as it appears online. Information and knowledge that aren’t digitized or widely shared online often never make their way into AI models.

Within that gap, Klein notes, the humanities play a vital role. “One of the roles of humanities scholars is to educate the public about how and why AI works the way it does,” she says. “You need to know who created the tools, who was speaking and what data went in.”

An example of this approach comes from a recent study Klein co-authored in The Lancet. The paper describes how AI software used by health clinicians to transcribe patient notes promises to ease a doctor’s note-taking burden — one of the most often cited causes of physician burnout. Klein says the tools would work a lot better if they were developed with more realistic use-cases in mind.

“Almost all the language transcription models we have now are trained on Midwest American accents,” she says. Patients and clinicians who don’t share that accent are transcribed less accurately. Many systems also assume only two people are speaking — doctor and patient — overlooking common scenarios where family members provide support or translation, Klein notes.

Inclusive design across disciplines leads to more accurate and equitable tools, Klein says. She calls this “humble AI” — narrow applications that aren’t glamorous but can improve people’s lives.

AI built by and for its users underpins the work of the Atlanta Interdisciplinary AI Network, which Klein leads. The group includes scholars from Emory, Georgia Institute of Technology and Clark Atlanta University working to bring AI innovations into language, politics and social activism, as well as health and science.

One project uncovers linguistic patterns in the writings of 19th-century abolitionists and suffragists. Their words reshaped America’s view of justice — broadening it beyond the legal to encompass fairness in everyday dealings with institutions. Studying how those definitions evolved, Klein says, helps today’s historians and political scientists better understand shifts in society and power.

Klein and her colleagues are also focusing on the 18th century to find untold or forgotten stories about the nation’s founding. Here, AI allows scholars to mine massive archives that would otherwise be impossible to fully explore. 

“Computational tools help researchers understand the social and communications networks of the founders, with 250 years of hindsight.”

—Dani Roytburg 25C

“The Founder’s Online dataset, with more than 180,000 documents, is too large for a human to process,” says Dani Roytburg 25C, a recent Emory graduate and former research assistant in Klein’s lab. “Computational tools help researchers understand the social and communications networks of the founders, with 250 years of hindsight.”

By tracing how language circulated through personal letters, Roytburg says that “we can locate how keywords traveled through correspondence to achieve prominence in the American Revolution vocabulary.” One of his goals was to use AI to understand better how the founders’ political affiliations shifted over the course of the Revolutionary War.

Roytburg values AI in research but is emphatic about its limits: AI does not replace scholarship or the need for deep historical knowledge, he says. Like physicians, historians can make and use these tools to extend the horizons of their informed analysis.

Barbara Krauthamer, a historian and dean of Emory College, adds that Emory’s efforts in applying technology and AI to the humanities are not just about merging data with discourse, but also about amplifying voices and narratives that might otherwise remain unheard. The result is innovation with a purpose: technology that preserves, rather than erases, the human story.

BUILDING GUARDRAILS AND VISION

At Emory, AI is more than an innovation. Emory faculty members are working to ensure the future of AI technology reflects human values. That means creating guardrails: systems of oversight, inclusive participation and bias testing that keep AI safe, fair and transparent.

Joe Sutherland, director of the Emory Center for AI Learning, puts it plainly: “Aligning AI with human values isn’t just a technical challenge. It’s fundamentally social. Can we create responsive guardrails that adapt to evolving societal values? Emory’s mission is to answer that challenge, pairing human insight with technological progress.”

The vision is expansive but grounded. In its best form, Sutherland says, AI could free people from repetitive tasks, giving them more time for creativity, inquiry and care. Emory’s commitment ensures that when those benefits arrive, they do so in ways that lift communities rather than fracture them.

Safdar adds that universities like Emory must lead in setting the standards. “AI use must align with Emory’s mission, principles and vision. We’ll always adhere to laws, regulations and policies, and we will need to develop new policies as the technology evolves,” he says. “We aim to foster transparency and empower people to use AI when it helps them — safely and responsibly.” 

Photos by Ben Knisely, Kay Hinton and Emory Photo/Video. Video courtesy of Advancement and Alumni Engagement.