Advancing Human-Centered AI Through Visual Data Analysis

When Dr. Chen He arrived at the International Computer Science Institute (ICSI) as a FARIA-ICSI International Research Fellow, she was looking for more than just a place to continue her research. She was seeking an environment where collaboration, interdisciplinary thinking, and real-world impact could shape the next stage of her work in information visualization and human-centered AI.

A postdoctoral researcher in the Department of Computer Science at the University of Helsinki, Dr. He earned her PhD in Computer Science in 2022 and has focused her research on how people interact with data through visual tools. During her time at ICSI, she worked under the mentorship of Dr. Marti Hearst from the UC Berkeley School of Information, a leading researcher in information visualization and information retrieval, whose work closely aligns with her own.

Designing AI systems that help people think, not just answer

At ICSI, Dr. He explored how large language models (LLMs) can support analytical reasoning through interactive data visualization. Rather than using AI simply to generate answers, her research focused on how AI can guide users through the process of discovering insights themselves.

“I designed a system that integrates LLMs into interactive visualization to guide users as they explore data and answer questions,” she says. “Instead of completing the task for the user, the system scaffolds their reasoning and encourages hands-on exploration.”

Her work builds on earlier research in visual analytics, but adds a new dimension by incorporating conversational AI into the analytical workflow. The prototype system was designed using a social constructivist learning approach, where the goal is to help users develop transferable analytical skills rather than rely on automation.

To evaluate the system, Dr. He conducted an exploratory study with 48 participants, comparing the LLM-assisted interface with a baseline system that offered only static guidance. The results showed that users with LLM support engaged in more systematic and comprehensive data exploration, suggesting that well-designed AI tools can strengthen human reasoning rather than replace it.

Working closely with Dr. Hearst helped refine both the technical and experimental aspects of the project.

“Marti’s guidance greatly strengthened my approach to experimental design, especially in ensuring realistic tasks and ecological validity,” Dr. He notes.

She also collaborated with members of Dr. Hearst’s group, including Chase Stokes, Jasmine Shih, and Samia Menon, whose feedback helped shape the direction of the research.

A community that extends beyond research

For Dr. He, one of the most meaningful parts of the fellowship was the sense of community she found at ICSI and Berkeley.

Through Dr. Hearst’s connection to the UC Berkeley School of Information, she joined monthly meetups with Ph.D. students and postdoctoral researchers, where conversations ranged from research ideas to academic career paths. She also participated in gatherings organized by the Berkeley International Office, where visiting scholars shared experiences and supported one another while adapting to life in a new country.

“These conversations broadened my view of research and academic life,” she says. “I also really enjoyed exchanging ideas and cultures with other visiting researchers at ICSI.”

The collaborative and welcoming atmosphere made it possible to pursue ambitious research while building lasting professional relationships.

“ICSI provided not only academic collaboration but also a supportive environment,” she reflects. “It created memories and connections that I will carry with me back to Finland.”

Looking ahead

Dr. He’s work on LLM-assisted visual data analysis reflects a growing shift in AI research, from building systems that replace human effort to designing tools that enhance human understanding. Her time at ICSI helped deepen that perspective, reinforcing the idea that the most powerful AI systems are those that help people think more clearly, explore more deeply, and learn more effectively.

As she continues her research career, the experience at ICSI stands as an example of how international collaboration, mentorship, and interdisciplinary thinking can drive new ideas at the intersection of visualization, machine learning, and human-centered AI.