|Shreepriya Shreepriya, Christophe Legras, Stéphane Clinchant, Jutta Willamowski|
|ACM Conference on Human Factors in Computing Systems (CHI), Online Virtual Conference (originally Yokohama, Japan), 8-13 May, 2021|
Recommender systems for runners primarily rely on existing running traces in an area. In the absence of running traces in an area, recommending running routes is challenging. This paper describes our approach to generating and proposing “pleasant” running tours that consider the runner’s standard preferences and their distance and elevation constraints. Our algorithm is an approach to solve the cold start recommendation problem in unknown places by mining available map-data. We implemented a prototypical smartphone app that generates and recommends pleasant running routes to evaluate our algorithm’s effectiveness. An in-the-wild user study was conducted, with 11 participants across three cities. We tested the correlation between what is defined as “pleasant path” by our algorithm and the user’s perception. The results of the user study show a positive correlation and support our algorithm. We also outline implications for the design of successful recommendation algorithms for runners.