THE PROJECT
React to This! is a ROSIE Lab research project about how people respond to non-verbal virtual agents. We built an annotated dataset to support training and evaluation of socially intelligent agent behavior.
The dataset captures multimodal reactions - facial expression, body language, and self-reported emotion - across different social scenarios, so future agents can respond in more appropriate human contexts.
MY CONTRIBUTION
I made significant contributions across multiple areas: created a descriptive/documentary video to communicate the research, designed and built the project website, edited the research paper, annotated data for analysis, and recruited participants and volunteers for data collection.
PROJECT LINK
rosielab.github.io/react-to-this/DESIGN CHOICES
Colour system
Typography
Tables and annotation structure
Dataset tables were structured for fast scanning: consistent column order, clear labels, and compact row spacing so coders could move quickly without losing context.
Grouping fields by modality (facial, body, self-report) made annotation review easier and reduced mistakes during quality checks.
Why these choices mattered
The goal was not visual novelty - it was clarity and trust for researchers, participants, and future readers of the dataset.
A calm colour system, readable typography, and structured tables made the project easier to use and easier to communicate.
DESIGN SHOWCASE
REFLECTIONS
Research as design
Working in a research lab taught me that defining the right question is harder - and more important - than building the answer. Designing the data collection process and the participant experience was as much a design problem as the website.
Making research accessible
The website and documentary weren't afterthoughts - they were the bridge between the paper and the people it could help. I learned how much communication design matters in academic work, where the default is to assume the audience will find you.