Collaborating Across Disciplines: Assistant Professor Lingyao Li

I am especially excited by the opportunity to collaborate across information science, computer science, health informatics and other domains while building a research program that bridges technical AI advances with human and societal needs.

  
Lingyao Li, who holds a PhD in Civil and Environmental Engineering, joined the College of Information Science after serving as an assistant professor at the University of South Florida and before that a postdoctoral researcher at the University of Michigan. His research focuses on how we can design artificial intelligence (AI) systems that are not only capable, but also trustworthy, adaptive and genuinely useful to people.

What brought you to the College of Information Science?

The College of Information Science stands out to me because of its interdisciplinary culture, collegial environment and the breadth of research that connects naturally with my work in large language models (LLMs), AI agents, human-AI interaction and AI for social good. I am especially excited by the opportunity to collaborate across information science, computer science, health informatics and other domains while building a research program that bridges technical AI advances with human and societal needs.

Before joining the University of Arizona, I was an assistant professor in the School of Information at the University of South Florida. Before USF, I was a postdoctoral researcher at the University of Michigan School of Information. My background is quite interdisciplinary, and over the years my research has gradually moved from civil engineering and data science toward LLMs and human-AI interaction.

What is your current research, and what most excites you about this work?

My current research focuses on how we can design AI systems, especially leveraging LLMs and AI agents, that are not only capable, but also trustworthy, adaptive and genuinely useful to people. One current focus is agent memory, such as how an AI agent can retain experiences, learn from interactions and use that history to provide more personalized and context-aware support. I am particularly interested in applying these ideas in health informatics, including patient-facing and simulated-patient settings, as well as in broader social-good applications.

What excites me most is the shift from static AI systems toward self-evolving agents that can learn from experience, use tools, collaborate with other agents and improve over time. I am especially interested in ideas such as recursive self-improvement, where an AI system can reflect on its own performance and iteratively improve how it reasons, plans and acts. Looking ahead, my group hopes to study both the technical mechanisms behind AI agents and the human-centered questions that come with them: What should an agent remember? When and how should it adapt? And how can we ensure that continued self-improvement remains useful, safe and aligned with human needs?

Tell us about your new research lab.

I am especially excited about launching the Sociotechnical AI Lab (SAIL) at Arizona and building new projects around LLM agents and human-AI interaction by addressing the sociotechnical challenges. My team is currently working on several projects involving long-term memory for AI agents, as well as applications in mental health and electronic health record (EHR) data. We are also beginning to explore self-evolving agents and recursive self-improvement. I am looking forward to collaborating with students and colleagues as these projects take shape.

Are you currently consulting in your field?

I have done a lot of application studies, although most of it takes place through interdisciplinary research collaborations rather than traditional consulting. I work with collaborators in health, computing, engineering and the social sciences to translate AI methods into real-world settings. For example, we have leveraged AI agents and LLMs for simulated patients and clinical decision support, cybercrime risks and disaster response. These collaborations help ensure that the technical questions we study are grounded in actual user and societal needs.
  

Tell us about your editorial and academic service.

I regularly review for major AI, natural language processing (NLP), human-computer interaction, health informatics and engineering venues, including ACL: Association for Computational Linguistics, ICLR: International Conference on Learning Representations, the Web Conference, AAAI: Association for the Advancement of Artificial Intelligence, Science Advances, npj Digital Medicine and JAMIA: Journal of the American Medical Informatics Association, among others. I have also served on conference program committees for AAAI and as an area chair for the Association for Computing Machinery’s Conference on Human Factors in Computing Systems (CHI). 

At my previous institution, I also served on a faculty search committee and several doctoral dissertation committees. Earlier in my career, I participated in international student support activities and community volunteering. Across these roles, I see service as a way to strengthen both the scholarly community and the environment in which students and colleagues can succeed.

What courses will you be teaching for the college, and what do you most enjoy about teaching?

I plan to teach Introduction to Generative AI, which will cover the foundations of modern generative AI, including LLMs, multimodal models and diffusion models, along with practical techniques such as prompt engineering, retrieval-augmented generation, fine-tuning and AI agents. The course will also emphasize how to evaluate these systems for reliability, safety and responsible use. The course is planned to be offered in Spring 2027.

What I enjoy most about teaching is the interaction with students. Their questions, perspectives and even the ways they approach problems often help me see familiar ideas from a different angle and refresh my own thinking. I also find it very rewarding to watch students gradually become more confident and independent. My goal is not only to help them learn the material in class, but also to inspire them to become self-motivated explorers who are curious enough to ask their own questions and continue learning beyond the classroom.

How do you bring your research into your teaching?

I try to make the classroom reflect the way research actually happens. In an LLM course, students will not only learn concepts such as prompting, retrieval-augmented generation, fine-tuning and agentic workflows; they will also evaluate hallucination, uncertainty, bias, privacy and other real deployment risks. I also bring examples from my own research in human-AI interaction, health informatics, social media analysis and AI agents so students can see how these methods are applied across different contexts.

I use project-based assignments and research papers to help students practice the full research process from asking questions and building systems to evaluating results and communicating limitations. Rather than simply reproducing code, I want students to understand why a method works, when it may fail and how to improve it. This connection between research and teaching keeps the material current and helps students see themselves as active contributors to the field, not just users of AI.
  

How do you engage with students to foster their academic and professional growth?

I engage with students by trying to create a supportive and responsive environment both inside and outside the classroom. In class, I use discussions, case examples and different types of assignments to encourage students to actively work through ideas rather than simply listen to lectures. I also pay close attention to their questions and feedback so I can adjust my teaching and provide additional examples or guidance when needed. 

Outside the classroom, I try to be accessible when students need advice about research, assistantship applications or other academic and professional opportunities. I see mentoring as an important part of teaching, and I want students to feel comfortable reaching out when they are trying to figure out their next step. I hope to help them build not only technical skills, but also the confidence and independence to make their own decisions. One comment I have especially appreciated from students is that they find me very accessible outside the classroom and responsive when they need help or guidance.

Beyond research and teaching, what are your passions?

I love cooking and trying new foods. Music is another long-time passion of mine. As an undergraduate, I sang in a campus band and won a Top 10 Campus Singer award at my university. Although I haven’t practiced singing for quite a while, I still really enjoy listening to music. I also studied drawing for about seven years, which remains another creative interest of mine. When it comes to sports, I’m a big fan of pickleball and tennis, and I enjoy playing both whenever I have the chance.

What advice do you have for InfoSci students?

Following emerging topics and trying new techniques can be a useful way to discover questions that genuinely feel worth pursuing. From my own experience moving from engineering to AI, some of the most meaningful directions often come from staying curious and being willing to explore challenges. Information science changes quickly, so I believe one of the most valuable habits is becoming a self-motivated explorer by reading broadly, building things, testing ideas, asking questions and continuing to learn beyond the classroom.

Another lesson I have learned is to try first rather than overthink too much at the beginning. New ideas can feel difficult or uncertain, but taking the first step often makes the path clearer. Challenges are part of the process, and sometimes the best way to understand whether an idea is worth pursuing is simply to start working on it.
  

Learn more about Lingyao Li on his faculty page, or explore ways you can support the dynamic, student-invested faculty of the College of Information Science.