A team led by Johns Hopkins Computer Science alumna Isabel Cachola, Engr ’26 (PhD), has received a Best Paper Award for its work, “Information Satisfaction: A Reader-Centered Axis for Summarization Evaluation,” at the 2026 ACM Conference on Human-AI Complementarity and Alignment (HCOMP), held September 27–30 at Virginia Tech. The HCOMP community studies and designs systems that combine the complementary strengths of human and artificial intelligence to achieve outcomes neither could achieve alone in ways that are ethical, safe, and intentional.
As a PhD student, Cachola was a member of the Center for Language and Speech Processing and was advised by Mark Dredze, a John C. Malone Professor of Computer Science and the director of the Johns Hopkins Data Science and AI Institute. Cachola is now an assistant professor of computer science at St. Edward’s University in Austin, Texas, where she develops AI-based systems that make scientific discovery accessible while focusing on the needs of end users.
Along with co-authors Will Walden and Reno Kriz, both research scientists at the Human Language Technology Center of Excellence, and Dredze, Cachola sought to determine how useful computer-generated summaries actually are to users—for example, a scientist studying vaccine development needs different details summarized than a family doctor reading about the same research.
The team tested whether popular evaluation tools can detect meaningful changes in a summary’s information content and asked human experts to judge if the resulting summaries met their real-world informational needs. The researchers found that both traditional metrics and newer AI-based ones fell short, revealing a real gap between what they were measuring and what readers actually care about.
“When we’re building generative systems, it’s important that we consider who is using our system and whether we are truly meeting their individual needs,” says Cachola.