Chenglin Li 李承霖

Abstract. Two studies with 56 participants examine how visual context misalignment in instructional videos affects physical task performance and user perception. We compare in-context instructional videos (ICON) with conventional videos, then isolate four visual attributes: task object intrinsics, task object state, environmental context, and observational context. Aligned videos improve completion quality and speed, while individual misalignments degrade performance.

Abstract. RAVEN enables blind and low-vision people to query and modify virtual 3D scenes in real time using natural language. Evaluations with eight blind and low-vision participants and six Unity developers explore how conversational programming can support personalized accessibility, along with the challenges of reliability, transparency, and trust.