Evaluation of AI-based Feedback System for Reducing Sidewalk Riding by Shared e-scooter Users

Abstract

The objective of this study was to assess the efficacy of using AI-based auditory feedback and speed limitations on shared e-scooters equipped with computer vision sensors to reduce sidewalk riding. Spin, a US-based micromobility company, provided data from 100 e-scooters in Santa Monica, California: half with activated feedback systems and half with deactivated systems. From November 23, 2022, to February 14, 2023, 488 trips were recorded within Santa Monica. Empirical cumulative distribution function (ECDF) plots and Kolmogorov-Smirnov tests indicate that feedback and speed limitations induced a statistically significant reduction in the fractions of trip time and distance that were spent on sidewalks, and in the length and duration of individual segments of sidewalk riding. The feedback group spent 22% less time and 20% less distance on sidewalks compared to the no-feedback group. To assess whether the feedback decreased the likelihood of choosing the sidewalk as the next surface when the rider is on the street or bike lane, we used a binary logistic regression model. The models' results revealed a statistically significant association between receiving feedback and a reduced inclination to choose the sidewalk as the next surface. These results show that feedback from using onboard cameras and artificial intelligence systems that identify roads, bike lanes, and sidewalks can alter e-scooter users' decisions on where to ride, potentially reducing conflicts between pedestrians and scooter riders and increasing compliance with city ordinances.