A Qualitative and Quantitative Approach to Studying the Public Mood (Zeitgeist): What Frightens British People About Islam in the UK?

This article introduces a hybrid qualitative–quantitative methodology for studying public sentiment at scale. It does not attempt a comprehensive study of British public sentiment toward Islam, but rather introduces a model illustrated through a preliminary case study that enables us to parse public sentiment about Islam in the United Kingdom into previously undetectable categories of concern. Moving beyond keyword-based approaches to analyzing online discourse, we trained a machine-learning classifier, developed in collaboration with a domain expert and validated at 89% accuracy (F1 = 0.82), to detect expressions of concern about Islam in the UK independent of specific vocabulary. The model captured coded language, sarcasm, and indirect allusion that models based on fixed search terms miss. Applied to more than 11,000 posts on X (formerly Twitter), the LLM model classified public concern about Islam into thirteen distinct thematic domains (grievances), including the perceived spread of Sharia law in the UK, grooming gangs, the presumed transformation of national identity, and antisemitism. Rather than collapsing public concern into a single, undifferentiated category of “Islamophobia,” the model produces a rich dataset that, when analyzed with an LLM-based application, distinguishes between distinct categories of unease. A case study of rhetoric directed at Keir Starmer and the Labour Party demonstrates how these categories of discourse themes can be tracked over time. By enabling continuous, granular monitoring, the approach offers scholars and policymakers a foundation for distinguishing legitimate grievances from prejudice.