What are people saying in their communities?

Across the analysis, one theme in particular kept resurfacing: people's engagement with politics is predominantly local and personal. People are consistently more concerned by potholes, grocery prices and antisocial behaviour that they personally witness than any wider policy. This is reflected in our finding that dominant issues were transport, public services, the environment, crime, and businesses. Together, these accounted for over half of all comments.

This shows how people are by far more interested in what they see, feel, and experience in their daily lives than national rhetoric or abstract political debates. And they often struggle to connect the two, unsure how national politics relates to their local reality, reflecting generally poor communication by political bodies on what powers they have and the reasons behind their decisions. This disconnect has consequences, leading to widespread disengagement, hostility and a sense that politics is distant and that all politicians are inherently corrupt.

Add background content here.

The data in this dashboard was categorised using a Large Language Model (LLM), which analysed posts and discussions from local Facebook groups and organised them into broad political issues, alongside more detailed subcategories within each issue area. Rather than relying on predefined labels or manual coding frameworks, the model identified recurring themes organically from the data itself, allowing patterns in conversation to emerge naturally.

Using an LLM for both theme generation and classification helps reduce the risk of introducing individual researcher bias into the analysis. To ensure consistency and reliability, the model’s outputs were tested against human validation exercises, where samples of posts were independently reviewed and compared with the AI classifications.

Political discussion is inherently messy, overlapping, and context-dependent. Many conversations touch on multiple issues at once - for example, immigration discussions may also involve housing, the economy, or public services. As a result, there is significant overlap between some categories, which is to be expected. The categorisation framework should therefore be understood as a way of identifying dominant themes and patterns within discussion, rather than imposing rigid or mutually exclusive labels onto political conversations.

Campaign Lab is a collaborative network of progressive campaigners, researchers, technologists, and data scientists working to better understand what makes campaigning effective in the UK. We bring together people from across politics, tech, and data science to test ideas, develop practical tools, and build evidence about what actually drives impact on the ground.

Campaigning depends on the dedication of thousands of volunteers and organisers every year, but much of that work still relies on assumptions, habits, and inherited practices rather than robust evidence. We believe campaigns should have access to better research, better data, and better ways of understanding what genuinely influences political outcomes. By running experiments, sharing findings, and building open tools, we aim to make campaigning more evidence-driven and collaborative. If you're interested in learning more about what we do, take a look at our website here.

This dashboard was created using Categorum.ai, a platform built to support the analysis of complex and nuanced datasets using advanced AI technology. Categorum is designed for researchers, campaigners, and organisations working with large amounts of qualitative information, helping users categorise responses, identify patterns, and explore data more efficiently without losing important context or nuance. Check it out here.

Local Facebook Groups · UK · Jul 2023 – Jul 2024