Q&A with Assistant Professor Leo Zhicheng Liu on Bridging AI and Human Insight

University of Maryland Assistant Professor of Computer Science Leo Zhicheng Liu focuses his research on the intersection of artificial intelligence, human-computer interaction and data visualization. His work aims to make data analysis and communication more intuitive for domain experts who may not have a technical background in computing. Liu leads the Human Data Interaction Group, affiliated with the Human-Computer Interaction Lab (HCIL), a cross-disciplinary research center that brings together faculty from computer science, information studies, and education.

In this interview, Liu discusses his early research experiences, his current projects, and his perspectives on collaboration, challenges and the future of human-centered data science.

Was there a defining moment that shaped your career path into computer science?

I think it’s really when I started working on an undergraduate research project in college. That was the first research-oriented project I worked on, and I really enjoyed it. I liked reading the literature, understanding what people had been doing so far, and then defining my own problem and coming up with solutions. That process—of identifying gaps and trying to address them—was very interesting to me. Because of that experience, I decided to pursue more research, which led me to apply for graduate programs and eventually a Ph.D.

Can you tell me about your current research focus and what drew you to this field?

My research focuses on synthesizing artificial intelligence with human-centered methods to help domain experts perform data analysis and data-driven communication more effectively. By domain experts, I mean people who are experts in their fields—scientists, business professionals, designers—who might not have a background in computer science or data analysis.

I’m interested in understanding how to design systems that can assist these users, combining computational methods with insights from human-computer interaction and visualization.

What projects are you currently working on?

I’m working on two main projects. The first explores how to automatically generate data visualizations or information graphics tailored to different users and their scenarios. For instance, a data analyst might want to visualize a dataset to gain insights, while a journalist might use visualization to communicate findings to a general audience. We aim to automate parts of this process to make the workflow more efficient.

The second project involves recommending strategies and workflows for data analysis. When users face complex datasets, they must decide which algorithms, models, or visualization techniques to apply. We’re developing systems that can suggest appropriate methods and combinations of tools to guide users through their analysis.

Both projects require a deep understanding of human needs and technical capabilities. We combine methods from machine learning, human-computer interaction, and visualization to create systems that are both intelligent and user-centered.

Are you affiliated with any lab at UMD?

Yes. Each faculty member typically leads their own research group. My group is called the Human Data Interaction Group, and we’re affiliated with the Human-Computer Interaction Lab here at UMD. The HCIL is a cross-departmental research center involving faculty from computer science, the iSchool, and the College of Education.

We hold regular meetings, brown-bag talks, and collaborative projects. It’s a vibrant environment that supports interdisciplinary research, which is essential for studying how humans interact with complex data and systems.

What is one challenge you’ve encountered in your research, and how did you approach it?

One challenge we’ve worked on involves enabling users to reuse existing data visualizations. Imagine you see an information graphic online and want to apply its design to your own dataset. To do that, the computer must first understand the structure and meaning of the original visualization.

Our approach involves developing a theoretical framework to identify the essential components of visualizations—what we might call primitives—that can be recombined in different ways. We then create algorithms and models that can deconstruct existing designs into these components. Ultimately, the goal is to build human-centered systems that turn existing visualizations into reusable templates.

It’s a complex process that requires both conceptual understanding and technical implementation.

How does your work connect to the broader computer science community or society?

Within computer science, my work contributes to developing new interfaces and representations that connect humans with data systems. It sits at the intersection of machine learning, databases, and human-computer interaction.

For example, visualization researchers and machine learning experts often collaborate to make complex models more interpretable. We also apply machine learning to simplify visualization design. Similarly, visualization and database researchers work together to ensure that data management systems can dynamically respond to user interactions.

From a societal perspective, the human-centered approach is key. My research aims to make it easier for professionals in different fields to interpret and communicate data-driven insights. One collaboration, for instance, involves working with cybersecurity experts to help them communicate network vulnerabilities to diverse stakeholders. In that sense, our work supports more effective decision-making and data communication across domains.

What inspired you to join the University of Maryland, and what have you enjoyed most so far?

UMD has a long-standing tradition in human-computer interaction research. The Human-Computer Interaction Lab is one of the oldest and best-known in the United States. When I first became interested in visualization as an undergraduate, many of the papers I read came from UMD. That early exposure influenced my interest in the field.

The location is also a major advantage. Being in the Washington, D.C., area provides access to many institutions, organizations, and collaborative opportunities. It’s an ideal environment for interdisciplinary work.

What advice would you give to students interested in your line of research?

In visualization and human-centered data science, it’s important to solve real-world problems and build tools for real users. But beyond that, I encourage students to think about how their findings can generalize beyond specific cases.

For example, after conducting user studies or domain-specific projects, we should consider how the lessons learned can inform broader theories or frameworks that others can apply. Balancing practical applications with theoretical insight allows research to have a lasting impact and transferability across domains.

—Story by Samuel Malede Zewdu, CS Communications