A new study by the Socially Intelligent Artificial Systems (SIAS) research group at the University of Amsterdam shows how political debates, news coverage and online interactions can collectively shape a discriminatory climate in the Netherlands. Commissioned by the Dutch State Commission against Discrimination and Racism, the study introduces a computational framework to measure these dynamics over time.
The SIAS team, consisting of João Fonseca, Gossa Lô, Erman Acar and Fernando Santos, analysed more than a decade of Dutch-language texts from three domains: parliamentary transcripts, national newspaper articles and comments on major Dutch news channels on YouTube. The analysis focused on seven legally recognised grounds for discrimination. The dataset includes millions of time-stamped texts and reactions.
Using an open-source pipeline, the researchers combined keyword analysis, sentiment analysis, large language model classification and time-series methods, including Granger causality. This approach enabled them to detect trends and cross-domain influences.
The findings show that discriminatory language is not confined to isolated online spaces. Online comment sections display increasing activity related to antisemitism, Islamophobia and discrimination based on origin. Since 2019, posts concerning origin and Islamophobia have risen on the YouTube channel of De Telegraaf, and in 2024 similar patterns appeared in the comment sections of NOS.
A key insight is the reciprocal influence between offline and online discussions. Parliamentary debates often precede shifts in discriminatory discourse on YouTube and in newspapers, while online discussions sometimes anticipate later political attention to sensitive topics. The mention of specific political parties in news coverage is statistically associated with subsequent increases in potentially antisemitic, Islamophobic and origin-related expressions in YouTube comments. This suggests that discriminatory narratives are becoming more embedded in mainstream public discourse.
The researchers argue that discrimination should be understood as a dynamic climate rather than a series of isolated incidents. They recommend integrated monitoring systems that combine parliamentary, media and online data to provide early warning signals of escalating discriminatory speech. The report also calls for clearer norms and behavioural guidelines within parliament to safeguard democratic standards.
Read also
