
Shlomo Zilberstein
Shlomo Zilberstein has long been fascinated by the foundational questions that define the promise and limitations of artificial intelligence—how intelligent behavior can emerge from computation and where its boundaries must be drawn.
With a lifelong passion for math, he was drawn to study computer science. “I was drawn to artificial intelligence because of the potential of computers to help us solve complex problems in a systematic way," he says. “I was interested in exploring the limits of what AI could achieve. Will artificial intelligence be able to do things better than humans? And how can intelligent systems reflect on their own thought processes so they can realize when they’re not on the right track and change course?”
Today, Zilberstein is a professor of computer science at UMass Amherst and director of the Resource-Bounded Reasoning Lab, where his research explores the computational foundations of automated reasoning and action. He studies how intelligent systems can operate independently given uncertainty in the environment, missing information, and limited computational resources.
But when Zilberstein first began studying AI, its prospects looked somewhat dim. “Back then, most problems felt impossible to solve. The available methods either didn’t work well or couldn’t scale,” he recalls. “But that’s what made it exciting—there was so much room to explore, and every breakthrough felt meaningful.”
Zilberstein attended graduate school in computer science at the University of California, Berkeley. He earned his PhD in 1993 during the second “AI winter,” a dark period when the development of AI encountered significant setbacks, as well as an economic downturn. He counts himself fortunate to have joined the faculty at UMass Amherst, an AI research powerhouse since the field’s early days. Zilberstein’s path to the academy was improbable in another way: His parents were Holocaust survivors who never had the opportunity to complete their own education. As the first in his family to attend college, the thought of becoming a professor felt daunting; at times, Zilberstein says he questioned his ability to make it. But his parents always had confidence in him, which gave him the courage to pursue an academic career and explore the big questions that captivated him.
Over the past three decades of his career, Zilberstein has made significant contributions to his field. His research has been recognized with numerous prestigious awards, including the 2025 ACM/SIGAI Autonomous Agents Research Award and the 2019 IFAAMAS Influential Paper Award. He is a fellow of the Association for the Advancement of Artificial Intelligence (AAAI) and the Association for Computing Machinery (ACM), and a recipient of the 2019 UMass Chancellor’s Medal.
These awards “recognize not only Professor Zilberstein’s intellectual brilliance but also his unwavering dedication to ethical and socially responsible AI development,” says Justin Domke, associate professor of computer science at UMass. “Beyond his own work, Zilberstein is a champion for knowledge sharing and collaboration within the AI community. His numerous roles as editor, conference chair, and board member demonstrate a commitment to fostering an inclusive and supportive environment for researchers across the globe.”
I was drawn to artificial intelligence because of the potential of computers to help us solve complex problems in a systematic way.
In recent years, Zilberstein’s research has expanded to consider how AI systems can make intelligent decisions when interacting with people and machines, given the inevitable uncertainties in any “open world,” whether physical or virtual. And though he began by exploring theoretical ideas, his research has been translated into real-world applications. Zilberstein often partners with industry on research and has had more than a dozen U.S. patents granted, along with many in other countries. Applications of his research include autonomous vehicles, which he has worked on in a partnership with Nissan, and space exploration systems like the Mars rover, which he worked on for NASA. Zilberstein has also collaborated with UMass colleagues on research on mobile robotics systems, such as robots that can deliver packages in an office or help clear items from the floor.
As AI becomes increasingly more powerful, Zilberstein has turned his focus recently to safety considerations. “I have developed a particular approach to safety that is based on metareasoning—a system reflecting on its own thought process—and self-awareness of competence so that a system can identify its own limitations and respond appropriately,” he says. “My philosophy about autonomy is that our goal is not to build systems that will do everything perfectly, reliably, and safely. Rather, they should be able to do a lot of things well but also recognize when they’re struggling and seek assistance while maintaining a ‘safe state.’”
Returning to a “safe state” may look very different depending on the application—picture the different actions an autonomous car, airplane, or Mars rover each needs to do to remain safe if they can’t continue operating as planned—but Zilberstein’s research seeks to develop fundamental mechanisms that would be applicable to many different applications. He is also studying potential unintended consequences of AI, such as how systems can be manipulated to act in harmful ways.
Zilberstein believes the future of AI lies in successful collaborations with humans. “I’m working on designing AI that is trustworthy and efficient and can be combined with the strengths of humans to enhance their everyday work in productive ways,” he says. “I believe this will potentially have a large impact on society.”
A seemingly simple example of this, which has been difficult to execute effectively, is creating automated personal assistants that can schedule meetings for a professional by automatically learning this individual’s scheduling preferences and conflicts. To be desirable, such an automated assistant would need to be safe, secure, and private.
Even as his research makes an impact in the real world, Zilberstein finds the most gratifying part of his work remains in finding answers to deep, fundamental questions that have never been answered before.
“It doesn’t happen a lot in a career, but there have been a few crisp results with clear implications that guide the future direction of the field,” he says. “I would like to be remembered for the lasting knowledge that I’ve contributed to the field of AI.”