Doctoral students honored for AI dissertations

UNIVERSITY PARK, Pa. — The third annual Penn State AI Awards recently recognized three doctoral students for their artificial intelligence (AI)-related dissertations that show how the systems and tools can be used or improved for the benefit of users.

Yuyang Deng, a postdoctoral scholar at Columbia University who graduated from Penn State in December 2024, won the award for Best Doctoral Dissertation on an AI-related topic for his project, “Collaborative Learning from Heterogeneous Sources with Provable Guarantees.”

“This award is such a great honor to me,” he said. “It recognizes that my research has impact and shows that hard work perseveres.”

The research project focused on users being able to train their own large language models on their own devices for the purpose of making AI systems more accessible, affordable, and personalized to fit any problem.

“There are millions of users that have their own data from their personal devices such as their computers, cell phones or smart watches,” Deng said. “What if there was a way where we could combine and use this data to learn how to and train a personalized AI model?”

For example, his research could be used in a hospital setting with doctors to optimize data and create predictive models for health care and diagnostic needs. The research could also be used in the development of autonomous driving vehicles by storing user data on directions, road conditions and more to create a decision-making model for rideshare navigation and driving.

“My research could be applied to many use cases,” Deng said. “The goal is to make AI systems more accessible to the public and make it so that everyone — researchers, doctors, the average person — can benefit from using it at no cost.”

Award honorable mentions, Teng Xiao and Pranav Narayanan Venkit, both Penn State informatics doctoral students who graduated this May, both pursued research avenues that focused on improving AI systems and how they interact with users.

Xiao’s project, “AI Learning and Alignment with Human Preferences,” aligned AI systems to better understand and respect human values such as honesty, fairness and safety. His study proposes machine learning algorithms that teach AI systems to connect more with these values on a faster time scale.

“Imagine you have a super smart robot that can write essays, solve math problems and chat like a human, but sometimes it says things that are confusing, wrong or even harmful. That’s a big problem — especially if we want to trust the systems in real life,” Xiao said. “We want to make powerful AI systems more aligned with human values, goals and expectations so that they don’t just act intelligently, but responsibly and safely.”

Similarly, Narayanan Venkit’s research, “Uncovering Automated Prejudices in Sociotechnical Natural Language Processing Systems,” is centered on building safer systems that interpret, understand and represent human language. The project also focused on defining and mitigating biases and connecting with users to gain insights on harm and biases. The team developed systems based on that feedback.

“In my first year of my doctoral program, I was trying to build a model that would decide if social media posts or comment sections were biased or not. I found that the terms that were picked up associated with different disabilities, and it was toxic language. I wanted to look into already existing models and make sure that they were safe to use. I wanted to create safer systems for everyone who uses them.”

The students hope that this domain of research continues to grow.

“I was surprised to find out I had been acknowledged with this award,” Narayanan Venkit said. “I was initially apprehensive when starting this research because of the distrust in AI systems and it wasn’t a popular field of study. Now, I am seeing how my research could play a huge role in peoples lives. People do care about having safer AI systems for various uses. I hope this award shows the next generation that this research matters.”

“Receiving this award is deeply meaningful to me, not just as recognition of my research but as validation of the core mission that drives my work: building AI systems that are aligned with human values and truly beneficial to society,” Xiao said. “I am grateful to my adviser, collaborators and mentors who have supported this journey. This award is not just a conclusion; it is a beginning.”

The Penn State AI Awards are coordinated and chaired by Vijaykrishnan Naryanan, Evan Pugh University Professor and associate dean for innovation in the Penn State College of Engineering. All awards are judged by a panel of Penn State Experts. Deng’s research adviser is Mehrdad Mahdavi, associate professor of computer science and engineering in the Penn State College of Engineering; Xiao’s research adviser is Vasant Honavar, Huck chair in biomedical data sciences and AI, professor and Edward Frymoyer Chair of Information Sciences and Technology in the Penn State College of Information Sciences and Technology (IST); and Narayanan Venkit’s was advised by Shomir Wilson, associate professor also in the College of IST.