Clustering Mouse Movement Behavior in Surveys Using ResNet Embeddings

Lisa Bondo Andersen, Tobias Wistuba, Felix Henninger, Ailin Liu, Sonja Greven, Frauke Kreuter

Keywords: Mouse tracking, survey behavior, cursor dynamics, ResNet embeddings, deep learning, clustering, behavioral patterns

TL;DR: We use ResNet to embed survey mouse movements and cluster them, exploring patterns across question types, trajectory features, individual differences, demographics, and self-reported task difficulty

Abstract: Mouse movement trajectories in online surveys has been shown to reflect question difficulty during online surveys. We explore the use of deep neural network embeddings to summarize these trajectories, using a ResNet-based architecture applied to time-normalized cursor paths. Clustering and UMAP visualization of these embeddings on a subset of the data reveal a combination of large, dense clusters and smaller, distinct subgroups, suggesting diverse movement patterns among respondents. These preliminary findings indicate that neural embeddings can capture meaningful structure in survey interaction behavior, providing a foundation for further investigation into individual differences and adaptive survey design.

Serve As Reviewer: ~Jason_Friedman1

Submission Number: 45