The Longitudinal Effects of Political Harmful Content on Trust in Politics Among Adolescents: a Data Donation Linkage Study

There is growing concern about the declining trust in politics among adolescents. Recent findings indicate that 71% of Belgian youth express very little or no trust in politics (Journée, 2024). Public discussions frequently point to a key aspect of young people’s daily lives as a potential explanation for this trend: their social media use. The algorithms of platforms such as TikTok and YouTube, which tend to push sensational short-form content through infinite scrolling, raise questions about whether young people are regularly exposed to material that could undermine their political trust. Existing literature suggests that general social media use (e.g., Lorenz-Spreen et al., 2023; Masoudnia et al., 2023) is associated with lower levels of trust in politics and government. However, we identify three important gaps in this body of research. First, most of these studies are broad in scope and rarely examine specific types of political content that might influence political trust. A comprehensive typology of potentially harmful political content (e.g., hate speech, conspiracy narratives, populist messaging, extremist material) is missing. Second, existing studies rely primarily on self-reported exposure measures, which offer limited insight into the actual prevalence or nature of the political content users encounter on social media platforms. Third, research focusing explicitly on adolescents is largely absent. Given that political attitudes and beliefs are especially malleable during adolescence (Andersen et al., 2020), and that these years are central to political socialization, it is therefore highly important. The present study is part of an FWO PhD fellowship (FWO is a leading funding organisation for research in Flanders, Belgium) and examines the longitudinal effects of exposure to political harmful content on political trust among adolescents. We apply a relatively new methodological approach, where we link survey data to a data donation. As such, we asked 760 adolescents (At T1) aged 16 to 18 from eight schools in Flanders (Belgium) to request their TikTok and YouTube watch and search histories and to complete a survey. This process is repeated three times over a period of eight months. In the first wave, 527 participants donated their data. T2 and T3 will take place in January and May, respectively. Using this dataset, we start with a semi-automatic content analysis. As such, political videos are first identified automatically through metadata screening using large language models (LLMs), after which they are manually coded for content type, characteristics, and source. This exposure measure is then linked to participants’ political attitudes spanning over three waves, allowing for within and between-person assessment. Through this methodology, we present a novel way to assess this critical question. We look forward presenting our results at the conference.