The Algorithmic Caricature: Auditing LLM-Generated Political Discourse Across Crisis Events

• URL: http://arxiv.org/abs/2605.12452v1
• Date: Tue, 12 May 2026 17:42:03 GMT
• Title: The Algorithmic Caricature: Auditing LLM-Generated Political Discourse Across Crisis Events
• Abstract summary: We construct a corpus of 1,789,406 posts across nine crisis events.<n>For each event, we compare observed discourse from social platforms with synthetic discourse generated for the same context.<n>Across events, synthetic discourse is fluent but population-level unrealistic.
• Score: 1.0666869578393299
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Large Language Models (LLMs) can generate fluent political text at scale, raising concerns about synthetic discourse during crises and social conflict. Existing AI-text detection often focuses on sentence-level cues such as perplexity, burstiness, or token irregularities, but these signals may weaken as generative systems improve. We instead adopt a Computational Social Science perspective and ask whether synthetic political discourse behaves like an observed online population. We construct a paired corpus of 1,789,406 posts across nine crisis events: COVID-19, the Jan. 6 Capitol attack, the 2020 and 2024 U.S. elections, Dobbs/Roe v. Wade, the 2020 BLM protests, U.S. midterms, the Utah shooting, and the U.S.-Iran war. For each event, we compare observed discourse from social platforms with synthetic discourse generated for the same context. We evaluate four dimensions: emotional intensity, structural regularity, lexical-ideological framing, and cross-event dependency, using mean gaps and dispersion evidence. Across events, synthetic discourse is fluent but population-level unrealistic. It is generally more negative and less dispersed in sentiment, structurally more regular, and lexically more abstract than observed discourse. Observed discourse instead shows broader emotional variation, longer-tailed structural distributions, and more context-specific, colloquial lexical markers. These differences are event-dependent: larger for fast-moving, decentralized crises and smaller for formal or institutionally mediated events. We summarize them with a simple event-level measure, the Caricature Gap. Our findings suggest that the main limitation of synthetic political discourse is not grammar or fluency, but reduced population realism. Population-level auditing complements traditional text-detection and provides a CSS framework for evaluating the social realism of generated discourse.

Related papers

• Elite Proxies, Algorithmic Bottlenecks, and Multi-Platform Information Cascades in the 2026 Iran War [0.0]
  This study tracks the transformation of President Trump's statements during the 2026 Iran War as they move from Truth Social into X-Twitter and Bluesky.<n>The results reveal pronounced behavioral divergence driven by socio-technical affordances.
  arXiv  Detail & Related papers  (2026-06-09T16:55:50Z)
• Anatomy of the Modality Gap: Dissecting the Internal States of End-to-End Speech LLMs [15.914430317382077]
  We analyze how speech and text representations evolve layer-by-layer.<n>We find that speech representations exhibit a broad cross-layer alignment band, attributable to the redundant nature of speech.
  arXiv  Detail & Related papers  (2026-03-02T06:21:43Z)
• Latent Topic Synthesis: Leveraging LLMs for Electoral Ad Analysis [51.95395936342771]
  We introduce an end-to-end framework for automatically generating an interpretable topic taxonomy from an unlabeled corpus.<n>We apply this framework to a large corpus of Meta political ads from the month ahead of the 2024 U.S. Presidential election.<n>Our approach uncovers latent discourse structures, synthesizes semantically rich topic labels, and annotates topics with moral framing dimensions.
  arXiv  Detail & Related papers  (2025-10-16T20:30:20Z)
• Generative Exaggeration in LLM Social Agents: Consistency, Bias, and Toxicity [2.3997896447030653]
  We investigate how Large Language Models (LLMs) behave when simulating political discourse on social media.<n>We construct LLM agents based on 1,186 real users, prompting them to reply to politically salient tweets under controlled conditions.<n>We find that richer contextualization improves internal consistency but also amplifies polarization, stylized signals, and harmful language.
  arXiv  Detail & Related papers  (2025-07-01T10:54:51Z)
• Talking Point based Ideological Discourse Analysis in News Events [62.18747509565779]
  We propose a framework motivated by the theory of ideological discourse analysis to analyze news articles related to real-world events.<n>Our framework represents the news articles using a relational structure - talking points, which captures the interaction between entities, their roles, and media frames along with a topic of discussion.<n>We evaluate our framework's ability to generate these perspectives through automated tasks - ideology and partisan classification tasks, supplemented by human validation.
  arXiv  Detail & Related papers  (2025-04-10T02:52:34Z)
• Towards Event Extraction from Speech with Contextual Clues [61.164413398231254]
  We introduce the Speech Event Extraction (SpeechEE) task and construct three synthetic training sets and one human-spoken test set. Compared to event extraction from text, SpeechEE poses greater challenges mainly due to complex speech signals that are continuous and have no word boundaries. Our method brings significant improvements on all datasets, achieving a maximum F1 gain of 10.7%.
  arXiv  Detail & Related papers  (2024-01-27T11:07:19Z)
• The Problem of Alignment [1.2277343096128712]
  Large Language Models produce sequences learned as statistical patterns from large corpora. After initial training models must be aligned with human values, prefer certain continuations over others. We examine this practice of structuration as a two-way interaction between users and models.
  arXiv  Detail & Related papers  (2023-12-30T11:44:59Z)
• ContextSpeech: Expressive and Efficient Text-to-Speech for Paragraph Reading [65.88161811719353]
  This work develops a lightweight yet effective Text-to-Speech system, ContextSpeech. We first design a memory-cached recurrence mechanism to incorporate global text and speech context into sentence encoding. We construct hierarchically-structured textual semantics to broaden the scope for global context enhancement. Experiments show that ContextSpeech significantly improves the voice quality and prosody in paragraph reading with competitive model efficiency.
  arXiv  Detail & Related papers  (2023-07-03T06:55:03Z)
• Revisiting Conversation Discourse for Dialogue Disentanglement [88.3386821205896]
  We propose enhancing dialogue disentanglement by taking full advantage of the dialogue discourse characteristics. We develop a structure-aware framework to integrate the rich structural features for better modeling the conversational semantic context. Our work has great potential to facilitate broader multi-party multi-thread dialogue applications.
  arXiv  Detail & Related papers  (2023-06-06T19:17:47Z)
• Collective moderation of hate, toxicity, and extremity in online discussions [1.114199733551736]
  We analyze a large corpus of more than 130,000 discussions on Twitter over four years. We identify different dimensions of discourse that might be related to the probability of hate speech in subsequent tweets. We find that expressing simple opinions, not necessarily supported by facts, relates to the least hate in subsequent discussions.
  arXiv  Detail & Related papers  (2023-03-01T09:35:26Z)
• Towards Debiasing Sentence Representations [109.70181221796469]
  We show that Sent-Debias is effective in removing biases, and at the same time, preserves performance on sentence-level downstream tasks. We hope that our work will inspire future research on characterizing and removing social biases from widely adopted sentence representations for fairer NLP.
  arXiv  Detail & Related papers  (2020-07-16T04:22:30Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.