Beyond the Lens: Detecting Bias through Eye-Tracking Measurements and Computer Vision Techniques

Abstract

This presentation investigates ethnocentric biases faced by refugee populations through empirical human-computer interaction (HCI) research. Using eye-tracking techniques, we analyzed participant responses to various images, revealing nuanced biases: participants favored certain refugees in complete images but shifted preferences based on text labels. This highlights how context influences perceptions of refugee identity.
Additionally, we explore the connection between authentic representations and increased support, demonstrating the potential of computer vision for social good. The study also employs art as a stimulus to untrain gender biases, emphasizing the importance of thoughtful representation in fostering inclusivity in discussions about marginalized communities.

Speaker Biography

Dr. Nada Attar earned her Ph.D. in Computer Science from the University of Massachusetts. She also holds an M.S. in Computer Science from Tufts University and a BSc. in Computer Science from King Abdulaziz University. Following her doctoral studies, she served as a Postdoctoral Research Fellow at Harvard University, specifically at the Mass. Eye & Ear Institute and the Schepens Eye Research Institute, Bowers Lab, in Boston, MA.
Dr. Attar’s research focuses on Computer Vision, Ethics in AI, and Human-Computer Interaction (HCI). She has contributed to various projects, including eye-tracking for automated vehicles, HCI for cognitive state measurement, and addressing gender and ethnic bias for social good. Her work also explores computer vision applications in climate change and coastal resilience, as well as ethical AI models that investigate biases in datasets, algorithms, and user behavior.

She is a recipient of the Mozilla Responsible Computing Challenge Award for her team’s work on CIRCLE: Cross-Campus Interdisciplinary Responsible Computing Learning Experiences.