ShowCAIS 2026 Chair, Co-Chair, organizing committee members, and CAIS Co-directors. (Photo Credit: Ashley Li)
In the world of artificial intelligence (AI), AI development can go in many directions, and experts see an increasing need for ethical AI and its use as a tool to drive social change.
From addressing mental health and homelessness to tackling the toughest issues facing veterans, as well as detecting wildfires and mitigating bias in job recruitment, AI has the potential to be used as a tool to address a wide range of social challenges.
At the USC Center for AI in Society (USC CAIS), researchers have long been on a mission to build AI for social good and foster a community of AI researchers across the university over the past decade.
Last month, the center hosted ShowCAIS 2026, its annual symposium where USC faculty and students from across the university present research using AI for the benefit of society.
The one-day event featured two keynote panels, poster sessions, oral presentations and networking opportunities where presenters exchanged ideas and discussed potential collaborations.
Students presenting AI research at the poster session (Photo Credit: Ashley Li)
The symposium was held March 27 at the Dr. Allen and Charlotte Ginsburg Human-Centered Computation Hall, opening its doors to more than 200 attendees and bringing together students and faculty from five USC schools.
Founded in 2016, USC CAIS is a decade-long partnership between the USC Suzanne Dworak-Peck School of Social Work and the USC Viterbi School of Engineering. As the center continues to expand and adapt to meet the evolving needs of today’s social challenges, this year marks its 10th anniversary.
This milestone is not only a testament to USC’s dedication to building AI for social good but also positions the university as a pioneer in developing technologies that benefit society in more ways than one.
ShowCAIS 2026 was hosted by chairs Swabha Swayamdipta, assistant professor in USC Viterbi’s Thomas Lord Department of Computer Science, and Angel Hwang, assistant professor at the USC Annenberg School for Communication and Journalism, with a joint appointment at USC Viterbi. Led by the two chairs, the symposium was made possible by an organizing committee of faculty, staff and student leaders from the USC Viterbi School of Engineering, the USC School of Advanced Computing and the USC Dworak-Peck School of Social Work.
USC CAIS is led by three directors: Bistra Dilkina, Dr. Allen and Charlotte Ginsburg Early Career Chair in Computer Science and associate professor in the USC Viterbi School of Engineering; Phebe Vayanos, who holds joint appointments in the Thomas Lord Department of Computer Science and the Daniel J. Epstein Department of Industrial and Systems Engineering; and John Blosnich, associate professor in the USC Suzanne Dworak-Peck School of Social Work; as well as three associate directors, Swayamdipta, Ajitesh Srivastava, assistant professor at USC Viterbi’s Ming Hsieh Department of Electrical and Computer Engineering; Lindsay Young, assistant professor of Communication at USC Annenberg.
The symposium’s three oral sessions were moderated by ShowCAIS 2026’s chairs, with the first one focusing on AI for clinical, mental and public health, the second on fairness in AI, and the final session on AI for the environment, wildfire monitoring and wildlife protection.
Dr. Allen Ginsburg and Charlotte Ginsburg attend ShowCAIS 2026 (Photo credit: Venice Tang)
Interdisciplinary Collaboration is Key to Driving AI for Social Impact
The opening keynote panel brought together leadership from engineering, computer science and social work to paint a picture of how USC has led — and will continue to lead — high-impact AI research.
As the panel’s moderator, Dilkina started by inviting panelists Gaurav Sukhatme, executive vice dean, director of the USC School of Advanced Computing and incoming interim dean of USC Viterbi’s School of Engineering; Shrikanth “Shri” Narayanan, USC University Professor and USC’s inaugural vice president for presidential initiatives; and Blosnich from USC’s School of Social Work, to share examples of their own research and AI’s impact on social causes.
Both Sukhatme and Narayanan shared perspectives on the computational and technical side of AI, each drawing examples from their past research. Sukhatme highlighted his work in marine robotics and AI for tackling environmental issues and solving water quality mysteries, while Narayanan outlined his interdisciplinary work using AI to model suicide ideation prediction and improve minority media representations. On the other hand, Blosnich offered a social science perspective, referencing his work on suicide prevention.
“What I really think is the secret sauce for USC’s innovative research is the interdisciplinary spirit,” said Dilkina. “And the effort that USC puts into the cultivation of collaborative efforts across campus.”
All three speakers offered varying perspectives on how AI can be used to tackle the issues society faces, but ultimately attributed these technologies and solutions to cross-field collaboration, especially as complex challenges require expertise from different disciplines.
Opening keynote panel moderator and speakers (left to right): Bistra Dilkina, Shrikanth Narayanan, John Blosnich, Gaurav Sukhatme. (Photo Credit: Venice Tang)
AI for Suicide Prevention and Mental Health
AI’s role in suicide prevention and mental health research was a major focus in both keynote sessions and widely discussed by presenters at this year’s symposium. Specifically, speakers from the second keynote panel, Srivastava, and Carl Castro, professor and director of Military and Veterans Programs at the USC Suzanne Dworak-Peck School of Social Work, led a discussion on “The Use of AI to Solve the Toughest Challenges Faced by Military and Veteran Populations.”
Suicide rates in the United States “have been going up … for the past generation,” pointed out Blosnich during the first keynote panel. With suicide being one of the most pressing problems facing the veteran population, Castro further highlighted the sharp rise in suicide rates in the U.S. Army since the mid-1980s.
Given this alarming trend, there is a need to find better solutions.
Blosnich emphasized that structural and physical interventions are highly effective and key to suicide prevention, citing data showing that 90% of individuals who survived jumping from the Golden Gate Bridge regretted it and did not attempt again.
From identifying individuals at risk to pinpointing the right opportunities to intervene, social scientists still face many roadblocks that make intervention difficult and often ineffective.
Students presenting AI research at the poster session (Photo Credit: Venice Tang)
One challenge is the difficulty humans face in recognizing patterns in complex, large-scale data. Data pattern recognition has mostly been limited to clinical settings, and social scientists often neglect “upstream signals,” such as interactions with nonclinical professionals who are not directly involved with high-risk individuals. This makes upstream intervention points invisible to researchers and leads to missed opportunities to stop tragedies from developing in early stages.
Another roadblock is the psychological load social workers face when manually processing and labeling data on suicides and violent deaths, often leading to “abstractor fatigue.” Many cases involve jarring scenarios and are really hard to read, making it difficult for social workers to maintain consistency in their labels. This then results in annotation inconsistencies and discrepancies that allow critical signals to be missed.
The complexity of micro behavioral patterns behind mental illnesses is another obstacle, as some indicators of risk are too subtle for a human observer to quantify. For example, in the first keynote panel, Narayanan discussed the challenge of analyzing behavioral data like eye movements and the “timing of your action,” which are difficult for humans to track at scale or over time with the precision needed to identify high-risk individuals.
Castro also pointed out a longstanding issue in mental health research: it often takes decades to produce solutions after a problem is identified. By the time researchers begin developing interventions, they are often “years too late” to address current conditions. This “research speed gap” makes traditional reactive models less effective in reducing suicide rates.
Students presenting AI research at the poster session (Photo Credit: Ashley Li)
These challenges highlight where AI can help fill critical gaps, helping researchers to “get in front of the curve.”
For example, while complex pattern recognition is difficult for humans, it is well suited for AI and large language models (LLMs). AI’s ability to analyze massive datasets makes it an ideal tool for processing large-scale narrative data, where models can identify “upstream” signals that humans may miss and reduce errors caused by abstractor fatigue. These insights can then serve as critical points for suicide intervention in early stages.
This approach can also help reduce the emotional burden on social workers, who are traditionally required to review and annotate graphic death reports.
In mental health research, AI’s ability to sift through data can not only accelerate research but also quantify behavioral patterns, significantly improving the accuracy of identifying risk factors and individuals at risk.
AI can also help shift suicide prevention from a reactive to a more proactive approach. LLMs can not only predict which individuals are at risk but also identify key factors that need to change to reduce that risk. For instance, both Blosnich and Castro highlighted that stronger relationships and human connections are critical protective factors. In military and veteran contexts, “unit cohesion” is especially important. These factors can be leveraged to help move individuals from high-risk to lower-risk profiles.
Full List of Best Poster and Presentation Winners
With a total of 11 oral presentations and 44 papers, students showcased their research throughout the symposium. Here is the full list of papers presented and below are the oral and poster presentation winners.
Oral Presentation Winners
- Jackson Trager — The Subjectivity of Respect in Police Traffic Stops: Modeling Community Perspectives in Body-Worn Camera Footage
- Jay Campanell — Forecasting Acute Child Malnutrition in Kenya
- Hannah Murray — Enabling Smart Sensor Placement for More Efficient Wildlife Monitoring
Poster Presentation Winners
- Daniel Ruiz — How Far Will They Go? Red Teaming Political Expressions of Large-Language Models
- Jaspreet Ranjit — Multi-Turn Interactions Reveal Hidden Safety Failures in Mental Health Language Model Use
- Siyi Zhou — The Trafficker’s Pitch: Detecting Deceptive Recruitment in Online Job Boards
Left to right: Yannis Yortsos, Bistra Dilkina, Charlotte Ginsburg, Dr. Allen Ginsburg, Kaci Silverman, Gaurav Sukhatme. (Photo Credit: Venice Tang)
To watch the oral presentations from the day, click here.
Published on April 15th, 2026
Last updated on April 21st, 2026