Crowdsourced Navigation in Mass Evacuation: A Lab Study on User Contribution

Crowdsourced navigation transforms users of transportation systems into live traffic sensors, generating real-time driving information that traditional map-based services do not capture. Despite its potential, the real-time contribution behavior of users in crowdsourced navigation remains largely unexplored. To address this gap, we developed a testbed enabling multiple human players to engage in an evacuation task with assistance from an AI guide, allowing them to report transient obstacles. A lab study with 32 participants was conducted, in which each participant completed four evacuation trials under different display configurations, varying the presence of contribution history and count information. Using a mixed-effects binary logistic regression model, we evaluated the effects of various motivating factors on contribution. Analysis of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{1, 0 1 2}$</tex> roadblock encounters revealed that contribution history information, prior trust, compliance with the AI guide, and roadblock exposure are key motivating factors. Future work may build on these findings to develop real-time behavior prediction models.