UBC Faculty Research and Publications

Communicating perceptions of pedestrian comfort and safety : Structural topic modeling of open response survey comments Bardutz, Emily; Bigazzi, Alexander York

Abstract

Purely quantitative or qualitative studies of perceived comfort and safety have different strengths and limitations related to scope, reliability, and generalizability. To pursue new insights about perceptions of pedestrian comfort and safety, this study uses the novel approach of Structural Topic Modeling to organize open-response comments into prominent topics, or themes, and explore the factors affecting the prevalence and content of topics. The study data come from a survey in which participants rated short video clips of pedestrians interacting with motor vehicles or bicycles at unsignalized crosswalks, and then provided optional open-response comments about each video. Modeling results identified six topics in the text data: (1) driver behaviour and pedestrian risk, (2) pedestrian risk assessment and mitigation, (3) pedestrians not waiting before crossing, (4) cyclists leaving the bicycle lane, (5) cyclists’ yielding necessity, and (6) and cyclists’ speed modifications. The prevalence of the topics was significantly affected by attributes of the interaction (vehicle type—bicycle or car—and its proximity), and attributes of the perceiver (travel habits and self-reported risk aversion). Topic prevalence was unrelated to participants’ comfort and safety ratings for the same interactions, indicating that topic modeling provides a fundamentally different source of information than quantitative severity ratings. The findings help to illuminate why pedestrian-cyclist interactions may attract disproportionate attention from the travelling public, despite their relatively lower risk than interactions with motor vehicle drivers.

Item Citations and Data

Rights

Attribution-NonCommercial-NoDerivatives 4.0 International