<p>This paper explores the relationship between public art and walking behavior using quantifiable approaches based on state-of-the-art machine-learning methods. The research team developed a preliminary method for using observational analysis and real-time object detection to study pedestrian engagement with public art. A custom machine-learning script was developed in Python based on open-source YOLOv3, an algorithm first described by Redmon et al. By building on Redmon et al.’s work, we increased the reliability and validity of our own custom script, which was tested on a range of walking conditions both at our site and using stock video footage of streets in other locales. Video recordings of pedestrian movements near sculptural artwork on Tufts campus captured by a low-tech security camera were run through the algorithm to generate path diagram images of passersby the artwork, enabling the research team to select videos where pedestrian behaviors of interest occur for empirical review. Research results evidence social walking behavior (groups of two or more) and proximity to walking path (adjacent to the sidewalk) are correlated with greater pedestrian engagement with artwork (looking at the art). These interdisciplinary methods successfully selected videos where pedestrians engaged with public artwork, verifying AI technologies are a promising research tool for studying walking behavior in a more data-driven way than has been done previously. Quantitative measures of pedestrian responses to public art can help urban planners and designers to promote art’s benefits including walkability, quality-of-life, and well-being.</p>