Overview
This PhD project will explore the application of large language models (LLMs) and vision language models (VLMs) to enhance the description, interpretation and visualization of geodemographic classifications. The research will investigate how generative AI can be leveraged to provide meaningful textual narratives and visual representations that characterise geographic areas, moving beyond traditional numeric profiling approaches.
The project bridges geodemographics, natural language processing, computer vision, and multimodal AI, with a particular focus on how VLMs and text-to-image models can generate representative visual archetypes that make geodemographic insights more accessible and engaging.
About this opportunity
This PhD will investigate how large language models and multimodal AI can enhance the interpretation and communication of geodemographic classifications. Central research questions address how LLMs can generate accurate, contextually appropriate narrative descriptions of geodemographic clusters; how combining LLMs with vision language models can improve the interpretability of complex demographic and geographic data; what limitations, biases, and ethical considerations arise when deploying generative AI for geodemographic characterisation; and how these approaches can be operationalised for policy and planning applications. Key research avenues include fine-tuning or prompt-engineering LLMs to produce coherent descriptions of geodemographic archetypes, evaluating output quality, accuracy, and reproducibility across different model architectures, and using VLMs to analyse satellite imagery, aerial photographs, and street-level data to extract semantic features such as urban density, vegetation, and built environment characteristics that can be linked to geodemographic profiles. A further strand explores the use of text-to-image models to generate representative visual archetypes for each classification, developing prompt strategies that accurately reflect cluster characteristics, evaluating stakeholder responses to AI-generated imagery compared with traditional statistical visualisations, and critically examining how such imagery might reinforce or challenge stereotypes. Methodological considerations span data integration combining traditional geodemographic variables with AI-derived interpretations from unstructured sources, bias assessment and mitigation across representational, algorithmic, and output dimensions, and ensuring explainability, transparency, and reproducibility in all outputs. Finally, the research envisages collaboration with policymakers and agencies in urban planning, public health, and transport to prototype communication tools and develop best-practice guidelines for the responsible use of generative AI in geodemographic research.
Further reading
Alex Singleton; Seth E. Spielman (2026). Computers, Environment and Urban Systems, 125, 102396. DOI: 10.1016/j.compenvurbsys.2025.102396
Alex Singleton; Seth E. Spielman (2024). EPJ Data Science, 13(1). DOI: 10.1140/epjds/s13688-024-00466-1
Alex Singleton; Dani Arribas-Bel; John Murray; Martin Fleischmann (2022). Computers, Environment and Urban Systems, 95, 101802. DOI: 10.1016/j.compenvurbsys.2022.101802