JANUARY 23, 2026
Ha Le
PhD Candidate
Khoury College of Computer Sciences
Northeastern University
About the Presentation: Accurate measurement of everyday behaviors is critical for research in human–computer interaction, ubiquitous computing, and personal health informatics, as it underpins many health tracking, assessment, and intervention systems. Yet most behavioral data collection studies still rely heavily on participants’ self-reports and manual activity annotations, which are time-consuming to produce, prone to recall errors, and often require substantial post hoc cleaning by researchers.
In this talk, I present two complementary systems, ACAI and GLOSS4HAR, that together explore how human–AI collaboration can improve the collection, validation, and refinement of activity annotations in free-living settings.
First, I introduce ACAI, a context-assisted Activity Annotation Interface that supports participants in efficiently labeling their daily activities. ACAI leverages system-generated activity suggestions and allows users to quickly accept or adjust labels while explicitly expressing uncertainty about temporal boundaries. Through a usability study and a two-week free-living deployment, we show that ACAI substantially reduces annotation time and perceived effort while improving data validity and fidelity compared to gold-standard activity recall methods commonly used in health research.
Building on the real-world dataset collected through ACAI, I then present GLOSS4HAR, a multi-agent, LLM-based system designed to assist researchers in cleaning and reconciling activity annotations. GLOSS4HAR mimics human sensemaking by triangulating passive sensing data with participant self-reports to detect inconsistencies, suggest corrections, and reconstruct fine-grained activity timelines. We demonstrate that GLOSS4HAR can both improve the quality of participant annotations and recover accurate minute-by-minute activity timelines from coarse or error-prone self-reports.
About the Presenter: Ha Le is a PhD candidate at the Khoury College of Computer Sciences at Northeastern University, advised by Prof. Stephen Intille and Prof. Varun Mishra. Her research sits at the intersection of Human–Computer Interaction, ubiquitous and wearable computing, and personal health informatics. She focuses on designing human-in-the-loop behavioral tracking systems that combine passive sensing with low-effort, user-driven feedback to improve personalization and real-world validity. Her work aims to help individuals more accurately track and reflect on everyday activities and behavioral patterns that matter to their health and wellbeing.
Ha’s research has been published in leading venues such as IMWUT/UbiComp, CHI, VIS, ASSETS, and PervasiveHealth. Her work explores multimodal sensing, uncertainty-aware self-reporting, and the use of AI/LLMs to support high-quality behavioral labeling and recall. She received her bachelor’s degree in Mathematics and Computer Science from Gustavus Adolphus College and has conducted interdisciplinary research spanning HCI, mobile and wearables health, and personal health informatics.