Bridging Minds and Machines: Orson Xu’s Vision for Human-Centered AI in Healthcare

Mental health often manifests in daily behaviors, such as sleep patterns, social interactions, and mobility routines. Xuhai “Orson” Xu envisions harnessing AI to analyze data from the likes of smartphones and wearable devices, offering users insights into potential mental health concerns and enabling real-time interventions to support their well-being.

Xu, the new Assistant Professor of Biomedical Informatics at Columbia University, has engaged in groundbreaking research across prestigious institutions driven by a singular vision: to build the next-generation intelligent interface between people and technology, particularly in healthcare.

With roots in engineering and a passion for human-computer interaction (HCI), Xu’s work focuses on using artificial intelligence (AI) to model human behavior and improve mental and physical health outcomes. He is excited to continue that research at one of the nation’s top medical centers while also mentoring the next generation of AI researchers.

Pioneering Behavioral Modeling for Mental Health

Xu’s PhD journey at the University of Washington (UW) exemplified his innovative approach to HCI and healthcare. By collecting data from smartphones and wearable devices, Xu and his team analyzed behaviors like sleep patterns, social activities, and mobility to assess mental health states. His research demonstrated that these external behaviors often mirror internal conditions such as depression or anxiety.

“What set our work apart,” Xu explains, “was not just the scale of the data we collected, but the emphasis on testing the generalizability of our models.”

By using datasets from multiple years and institutions, including Dartmouth and Carnegie Mellon University, Xu’s team could evaluate whether AI models trained on one group could accurately predict outcomes for another. This robust methodology addressed a critical gap in AI research: ensuring that models perform well across diverse populations. While Xu’s mental health models are groundbreaking, he emphasizes the careful consideration required for implementing interventions.

“Mental health is high-stake,” he said. “We need to be really careful before deploying any intervention.”

Instead of directly targeting depression, Xu initially focused on related issues, such as smartphone addiction. His team developed real-time, adaptive interventions informed by psychological theories like self-affirmation, combining traditional frameworks with cutting-edge AI. He is driving preliminary work that aims to provide real-time intelligent interventions for users at the risk of depression.

Closing the Loop in HCI

Xu’s interest in research began during his undergraduate years at Tsinghua University in China. While exploring various fields, he discovered his calling in HCI—a domain where technology intersects with human needs. Brief internships at Carnegie Mellon University and MIT deepened his fascination, providing a glimpse into the potential of technology to improve lives. A fortunate meeting at CMU would help shape his research journey; Xu met his future PhD mentor, Anind Day, who helped guide his work at UW. After completing his doctorate, he returned to MIT to embark on his postdoctoral research.

“I found HCI the most exciting because it allowed me to bridge the gap between human thoughts and external behaviors,” Xu shares.

Xu’s doctoral work laid the foundation for his lifelong research mission: creating a two-way loop where human behaviors inform AI models, and AI, in turn, provides interventions to improve human well-being.

Academic Honors and Leadership

Xu’s journey is distinguished not only by his research contributions but also by numerous academic honors. As a doctoral student, Xu received the Distinguished Dissertation Award and was named the Innovation in Technology Award recipient, and he has earned multiple Distinguished Paper honors at top-tier conferences such as ACM CHI and ACM UbiComp.

Beyond his research, Xu’s leadership extended to his peers and the broader academic community. At both UW and MIT, he actively mentored younger researchers and contributed to initiatives promoting interdisciplinary collaboration. Xu credits his mentors and colleagues for fostering an environment that allowed him to thrive.

“I’ve been fortunate to work with incredible people who challenged and inspired me,” he reflects.

Building Trust in AI Systems

One of the biggest challenges is fostering trust in AI systems, which will be essential for widespread adoption. He identifies two complementary approaches: explainable AI and gradual adaptation.

“Explainable AI aims to provide transparency,” Xu says. “By showing users why a model made a particular prediction, we can help them decide when to trust—or not trust—the AI.”

Xu’s research has shown that providing clear, evidence-based explanations increases user adoption and encourages reflection. At the same time, Xu acknowledges that trust often grows organically over time, and it can only truly occur when the technology itself is right.

Academia as a Catalyst for Impact

Xu’s decision to join Columbia University was guided by its emphasis on interdisciplinary research and its proximity to the medical community. Initially envisioning his career in computer science, Xu’s trajectory shifted as his research delved deeper into health and medical domains.

“Being in a medical school allows me to access real-world health data and collaborate directly with doctors and patients,” he explains. “It’s a chance to bridge my expertise in HCI with pressing medical challenges.”

At Columbia, Xu will lead the SEA (Sense, Empower, and Augment) Lab and plans to advance his work in creating AI-driven tools that improve clinical decision-making and individual’s well-being. From developing dashboards for mental health clinicians to designing systems for detecting atrial fibrillation, Xu’s projects aim to transform how health data is collected, analyzed, and acted upon.

By integrating wearable device data with electronic health records (EHRs), Xu envisions a future where AI doesn’t just augment clinical workflows but also empowers patients and providers with actionable insights.

“AI algorithms and human-AI collaboration workflows are potentially transferable across domains,” Xu asserts. “Whether it’s psychiatry, dermatology, or cardiology, the goal is the same: to create systems that improve outcomes while maintaining human oversight.”

Xu possesses industry experience as well; he currently serves as a Visiting Faculty within Google Health, where he focuses on wearable data and large language models for personal health.

A Vision Rooted in Humanity

Xu’s work embodies the promise of AI: technology that adapts to human needs rather than the other way around. At DBMI, he is determined to make this promise a reality, combining rigorous research with practical applications.

“It’s about building future tools and systems that not only work well and bring real-world benefit, but are trusted and embraced by the people who use them,” Xu says.