Author Information
Ian Drumm, University of Salford, United KingdomSer-Huang Poon, University of Manchester, United Kingdom
Abstract
This paper presents a transparent and reproducible pipeline for modelling online political discourse through clustered opinion spaces derived from real Reddit post–comment pairs. Addressing the challenge of mapping increasingly fragmented online ideologies, we use large-scale orchestration of LLMs for ordinal feature scoring and k-medoids clustering with Gower distance to construct interpretable archetypes that capture ideological, moral, and stylistic variation in public debate. These clusters act as filters within retrieval-augmented generation (RAG), enabling synthetic comments grounded in authentic discourse rather than arbitrary persona assumptions. Case studies on tariff debates and stock-market discussion illustrate the method’s flexibility in identifying specific discursive tropes within polarized communities. The pipeline offers a scalable approach for studying online discourse, providing interpretable synthetic data and a controlled framework for examining how discursive styles contribute to digital polarization.
Paper Information
Conference: ACSS2026Stream: Computational Social Science
This paper is part of the ACSS2026 Conference Proceedings (View)
Full Paper
View / Download the full paper in a new tab/window
To cite this article:
Drumm I., & Poon S. (2026) Orchestrating Large Language Models to Construct Interpretable Opinion Spaces for the Analysis and Emulation of Political Discourse ISSN: 2186-2303 – The Asian Conference on the Social Sciences 2026: Official Conference Proceedings (pp. 729-745) https://doi.org/10.22492/issn.2186-2303.2026.56
To link to this article: https://doi.org/10.22492/issn.2186-2303.2026.56
Comments
Powered by WP LinkPress