User data are key to understanding the diversity of individual cycling needs. Especially previous route choices allow for a deeper understanding of user behaviour and therefore how cyclists pick their routes when aiming to visit points-of-interests over the course of their journey. Today, AI-based recommender systems can help to provide meaningful information for cyclists allowing them to plan the routes according to their needs and preferences.
Yet, AI-based algorithms and the recommendations they produce raise important fairness question. In the presented research, we distinguish between user-centred fairness and service-provider-centred fairness. Ultimately, we aim at contributing to the development of a multi-stakeholder fairness approach by determining criteria according to which a proposed route recommendation can be considered to be fair.
Following up on our workshop at CRBAM in Zurich last year, we are now able to report on the findings of an empirical study, based on qualitative interviews with stakeholders and intermediaries from the tourism industry. We found that proximity to the intended route and other user-centred criteria have been identified as relevant fairness criteria. Regarding service providers, the interviewed stakeholders consider it fair to reward efforts made to meet the needs of cyclists. Direct measures to counteract popularity bias and over-tourism are also recognised as fairness criteria, yet to a lesser degree. In addition, our research has looked into the wider fairness implications of platform-based services and digital infrastructures required for cycling route recommendations to be made available.
We intend to share the results of our research in an interactive style, encouraging our audience to engage in a discussion on cycling fairness and the ways in which AI-based recommender systems may contribute to enhancing it. CAMPING rules will be applied to secure just that.