Immigration, Social Media and Narrative: An LLM-led Analysis

Motivation and Methods

Immigration consistently ranks as one of the issues of greatest concern among UK voters¹. As Reform has gained momentum based on its messaging around this issue, being able to understand and respond to public opinion on this is an increasing priority. This article summarises findings from a larger report produced by Campaign Lab in 2023, in which we applied a Large Language Model (LLM) to classify 8,734 posts drawn from local Facebook groups connected to that year’s hotel protests, with the aim of characterising how immigration is discussed in online community spaces.

Local community Facebook groups are a valuable source of data for this kind of research. Because participants are not aware of being studied, the content is less subject to social desirability bias than survey or interview-based methods, making these spaces an ideal window into unmediated public sentiment. Applying an LLM to classify posts at this scale substantially reduced the researcher time required, enabling a broader dataset than manual analysis would have allowed. Automated classification also mitigates the subjective influence that can arise when human researchers construct or apply categories by hand.

This research addresses two core questions: what are people saying about immigration, and how are they saying it? We were particularly interested in identifying the dominant language, frames, and narratives present in this discourse.

Overview of Results

Overall, we found that there is a high level of direct animosity between people who are negative and positive about immigration, with ad-hominem attacks making up 17% of the comment dataset. Our key finding was that in addition to this, the way that people who are positive and negative about immigration talk about it fundamentally differently: those who are negative use emotional language, while those who are positive use moral language. This distinction can go some way to explaining why immigration is such a divisive and apparently irreconcilable topic – the two sides are not simply disagreeing on substance, but are operating in fundamentally different rhetorical registers.

Interestingly, while attitudes towards immigrants in the comment set were significantly more negative than positive, much of the hostility was directed towards the government, and associated concerns around fairness and access to public services, rather than immigrants themselves. Additionally, only 2.5% of the comments in our sample expressed concern about immigration threatening British identity or replacing natives. This suggests that the dominant affective driver of anti-immigration sentiment in these communities is not cultural anxiety but basic economic concern.

Advice for Campaigners

These findings carry several concrete implications for progressive campaigners. First, the data suggest that progressives may be making a tactical error by defaulting to moral arguments when responding to positions that are rooted in and articulated through emotional language. Engaging on moral grounds with an audience reasoning emotionally is unlikely to cut through; more effective persuasion may require meeting people in their own register.

Second, it is worth noting that only 2.5% of comments expressed concern about immigration threatening British identity or replacing natives. This is relevant for those thinking about the kinds of messaging that might shift the dial on immigration rhetoric; prejudice towards immigrants is largely underpinned by basic economic concerns and fears around access to public services. We can therefore advise that progressives speak to people on these grounds rather than attempting to engage morally on immigration.

As Reform currently leads in the polls, with a largely anti-immigrant platform, it is imperative that progressives come to grips with how to engage with people on immigration in ways that resonate with them. As debates on immigration flare up in communities, it becomes critical to recognise the necessity for political strategies that work.

This research indicates that current methods of constructing moral arguments on immigration are unlikely to cut through to those who are skeptical, but that persuasion can be possible if we are able to tap into emotional language.We hope that the research and recommendations in this report are helpful in understanding and combatting anti-immigration rhetoric, a monumental task which has the potential to be key in keeping Reform out of government in the near future.

The full report is available below and is strongly recommended for any progressive campaigner or researcher working in this area.