Computer Science Assistant Professor Stevie Chancellor named Novus Faculty Impact Fellow

Inaugural award to advance research on safeguarding mental health outcomes using AI

MINNEAPOLIS / ST. PAUL (06/03/2026) — University of Minnesota Department of Computer Science & Engineering Assistant Professor Stevie Chancellor has been named a recipient of the Novus Faculty Impact Fellowship for her groundbreaking work in human-centered AI to improve mental health outcomes online.

The Novus Faculty Impact Fellowship was made possible by a generous gift from College of Science and Engineering (CSE) alumni John Dexheimer and Carla Haugen. It is awarded to exceptional faculty members to recognize those making outstanding contributions with a distinct emphasis on translational opportunities, commercial impact, and connections to data science or Artificial Intelligence. As the first-ever recipient of this award, Chancellor will serve a three-year term as the inaugural fellow, receiving discretionary funds to advance her high-impact research.

“Being named a Novus Faculty Impact Fellow is an incredible honor, and it comes at a critical tipping point for AI safety,” Chancellor said. “We are moving past the phase of just identifying algorithmic harms; this fellowship gives us the momentum to actively build human-centered solutions. It provides the crucial resources my lab needs to design AI architectures that actually empower users, give them back their agency and structurally prioritize mental health safety. I am truly so grateful to John Dexheimer and Carla Haugen for making this possible.”

Chancellor’s research builds and evaluates human-centered AI for mental health in online interactions. By analyzing digital trace data from millions of social media interactions and chatbot dialogues, her team develops novel AI architectures that can accurately model mental health behaviors, even when working with small or complex multi-modal datasets. Guided ethically by the lived experiences of communities and stakeholders, her work specifically targets the dangers of algorithmic recommendation systems for social media and unsafe chatbot design. 

A major focus of Chancellor’s recent work addresses how personalized social media platforms can trap users in harmful content loops that system participants have described as “runaway trains” and “dopamine slot machines.” To counter this, Chancellor’s lab is developing a novel AI architecture featuring a "choose your own adventure" style algorithm. This system handles passive content recommendation while actively handing agency back to the user by nudging them to take breaks, giving them explicit control over their feeds, and encouraging open communication with the AI when something feels wrong.

With the backing of the Novus Faculty Impact Fellowship, Chancellor’s group will also expand their critical safety evaluations of AI mental health chatbots. Her team’s previous research exposed alarming vulnerabilities in standard chatbot guardrails, revealing instances where systems ignored delusional behavior or reinforced stigma. Her upcoming research will investigate the underlying mechanisms that trigger these unsafe dialogue failures, aiming to build airtight guardrails into next-generation systems.

About Dr. Stevie Chancellor

Dr. Stevie Chancellor joined the Department of Computer Science & Engineering at the University of Minnesota in 2021 as an assistant professor. She received her MA (2014) in communication, culture, and technology from Georgetown University and her PhD (2019) in human-centered computing from the Georgia Institute of Technology. She has received multiple Best Paper awards and Honorable Mentions for her publications. Her insights on the intersection of AI safety and mental health are frequently cited in national media, including The New York Times, The Washington Post, and The Atlantic.

Learn more about Chancellor’s work at her personal website and watch her 2025 TED Talk. 

Story by Erin Edlund