Contesting personalized recommender systems: a cross-country analysis of user preferences

Abstract

Very Large Online Platforms (VLOPs) such as Instagram, TikTok, and YouTube wield substantial influence over digital information flows using sophisticated algorithmic recommender systems (RS). As these systems curate personalized content, concerns have emerged about their propensity to amplify polarizing or inappropriate content, spread misinformation, and infringe on users’ privacy. To address these concerns, the European Union (EU) has recently introduced a new regulatory framework through the Digital Services Act (DSA). These proposed policies are designed to bolster user agency by offering contestability mechanisms against personalized RS. As their effectiveness ultimately requires individual users to take specific actions, this empirical study investigates users’ intention to contest personalized RS. The results of a pre-registered survey across six countries – Brazil, Germany, Japan, South Korea, the UK, and the USA – involving 6,217 respondents yield key insights: (1) Approximately 20% of users would opt out of using personalized RS, (2) the intention for algorithmic contestation is associated with individual characteristics such as users’ attitudes towards and awareness of personalized RS as well as their privacy concerns, (3) German respondents are particularly inclined to contest personalized RS. We conclude that amending Art. 38 of the DSA may contribute to leveraging its effectiveness in fostering accessible user contestation and algorithmic transparency.