arXiv subjects

Christopher Barrie

Publications and source records attributed to Christopher Barrie.

8 recordsLinked to original sources

arXiv2026

LLM Recommenders Produce Demographic Inequality Through Content Preferences

Large language models (LLMs) are increasingly being integrated into systems that rank and recommend online content. We examine how this shift can produce demographic inequality in whose voices are surfaced. Using 10,000 Twitter/X posts written by 1,000 survey respondents and linked to verified self-reported characteristics across nine demographic attributes, we conduct a controlled recommendation experiment across three LLM providers, six recommendation objectives, and five levels of contextual information. By varying whether models receive post text alone, platform metadata, or explicit author demographics, we isolate three pathways to unequal recommendation: preferences for demographically patterned content, reliance on unequal engagement signals, and direct use of disclosed identity. We find that demographic inequality emerges from text alone. Models consistently favour longer, more negative, and more opinionated posts, particularly those concerning news and social issues, producing systematic differences in the demographic composition of recommended authors. Platform metadata generally amplifies these inequalities, while explicitly disclosing authors' demographic characteristics produces little additional change. Recommendation objectives mainly affect the magnitude of these patterns: controversial framing intensifies both content and demographic skew, whereas popularity-oriented framing dampens it, although for some attributes - most notably age - the objective can also reverse which group is favoured. LLM-based recommenders can therefore reproduce demographic inequality by proxy, not through explicit demographic targeting, but through preferences for content and engagement signals already structured by social inequality.

cs.CY↗

arXiv2026

Political Sorting Can Drive AI Models Apart Through User Feedback

Large language models are rapidly becoming an important source of political information. This raises a fundamental question: will AI systems support a shared basis for political knowledge, or lead different political groups to rely on increasingly different models? Political sorting can drive model fragmentation if three conditions hold: politically different users select into different models, learning from user feedback pushes those models apart politically, and the resulting differences shape subsequent model choices. We call this self-reinforcing process the Centrifugal Alignment Spiral. We study its components in three steps. First, we draw on a human experiment showing that political identity predicts model choice. Second, we fine-tune language models on synthetic feedback reflecting predominantly Democratic or Republican preferences. Across five independent runs per model family, paired models diverged on 12-41% of unseen survey questions with large partisan gaps, and in every run the differences moved in the expected political direction; for some models, differentiation extended even to issue areas excluded from training. Pooling feedback across groups instead suppressed divergence. Third, an empirically anchored agent-based model shows what follows when political sorting and model adaptation operate together: models attract politically distinct audiences, learn from them, and diverge further. User feedback can therefore turn political sorting among AI users into durable differences between the models on which they rely for political information.

cs.CY↗

arXiv2026

Prompt Stability Scoring for Text Annotation with Large Language Models

Researchers are increasingly using language models (LMs) for text annotation. These approaches rely only on a prompt telling the model to return a given output according to a set of instructions. The reproducibility of LM outputs may nonetheless be vulnerable to small changes in the prompt design. This calls into question the replicability of classification routines. To tackle this problem, researchers have typically tested a variety of semantically similar prompts to determine what we call ``prompt stability." These approaches remain ad-hoc and task specific. In this article, we propose a general framework for diagnosing prompt stability by adapting traditional approaches to intra- and inter-coder reliability scoring. We call the resulting metric the Prompt Stability Score (PSS) and provide a Python package \texttt{promptstability} for its estimation. Using six different datasets and twelve outcomes, we classify $\sim$3.1m rows of data and $\sim$300m input tokens to: a) diagnose when prompt stability is low; and b) demonstrate the functionality of the package. We conclude by providing best practice recommendations for applied researchers.

cs.CL↗

arXiv2026

Polarization by Default: Auditing Recommendation Bias in LLM-Based Content Curation

Large Language Models (LLMs) are increasingly deployed to curate and rank human-created content, yet the nature and structure of their biases in these tasks remains poorly understood: which biases are robust across providers and platforms, and which can be mitigated through prompt design. We present a controlled simulation study mapping content selection biases across three major LLM providers (OpenAI, Anthropic, Google) on real social media datasets from Twitter/X, Bluesky, and Reddit, using six prompting strategies (\textit{general}, \textit{popular}, \textit{engaging}, \textit{informative}, \textit{controversial}, \textit{neutral}). Through 540,000 simulated top-10 selections from pools of 100 posts across 54 experimental conditions, we find that biases differ substantially in how structural and how prompt-sensitive they are. Polarization is amplified across all configurations, toxicity handling shows a strong inversion between engagement- and information-focused prompts, and sentiment biases are predominantly negative. Provider comparisons reveal distinct trade-offs: GPT-4o Mini shows the most consistent behavior across prompts; Claude and Gemini exhibit high adaptivity in toxicity handling; Gemini shows the strongest negative sentiment preference. On Twitter/X, where author demographics can be inferred from profile bios, political leaning bias is the clearest demographic signal: left-leaning authors are systematically over-represented despite right-leaning authors forming the pool plurality in the dataset, and this pattern largely persists across prompts.

cs.SI↗

arXiv2025

Computational Turing Test Reveals Systematic Differences Between Human and AI Language

Large language models (LLMs) are increasingly used in the social sciences to simulate human behavior, based on the assumption that they can generate realistic, human-like text. Yet this assumption remains largely untested. Existing validation efforts rely heavily on human-judgment-based evaluations -- testing whether humans can distinguish AI from human output -- despite evidence that such judgments are blunt and unreliable. As a result, the field lacks robust tools for assessing the realism of LLM-generated text or for calibrating models to real-world data. This paper makes two contributions. First, we introduce a computational Turing test: a validation framework that integrates aggregate metrics (BERT-based detectability and semantic similarity) with interpretable linguistic features (stylistic markers and topical patterns) to assess how closely LLMs approximate human language within a given dataset. Second, we systematically compare nine open-weight LLMs across five calibration strategies -- including fine-tuning, stylistic prompting, and context retrieval -- benchmarking their ability to reproduce user interactions on X (formerly Twitter), Bluesky, and Reddit. Our findings challenge core assumptions in the literature. Even after calibration, LLM outputs remain clearly distinguishable from human text, particularly in affective tone and emotional expression. Instruction-tuned models underperform their base counterparts, and scaling up model size does not enhance human-likeness. Crucially, we identify a trade-off: optimizing for human-likeness often comes at the cost of semantic fidelity, and vice versa. These results provide a much-needed scalable framework for validation and calibration in LLM simulations -- and offer a cautionary note about their current limitations in capturing human communication.

cs.CL↗

arXiv2025

Emergent LLM behaviors are observationally equivalent to data leakage

Ashery et al. recently argue that large language models (LLMs), when paired to play a classic "naming game," spontaneously develop linguistic conventions reminiscent of human social norms. Here, we show that their results are better explained by data leakage: the models simply reproduce conventions they already encountered during pre-training. Despite the authors' mitigation measures, we provide multiple analyses demonstrating that the LLMs recognize the structure of the coordination game and recall its outcomes, rather than exhibit "emergent" conventions. Consequently, the observed behaviors are indistinguishable from memorization of the training corpus. We conclude by pointing to potential alternative strategies and reflecting more generally on the place of LLMs for social science models.

cs.CL↗

arXiv2022

Did the Musk Takeover Boost Contentious Actors on Twitter?

Twitter has been accused of a liberal bias in its account verification and content moderation policies. Elon Musk pledged, after his acquisition of the company, to promote free speech on the platform by overhauling verification and moderation policies. These events sparked fears of a rise in influence of contentious actors -- notably from the political right. In this article, I use a publicly released list of 138k Twitter accounts that purchased blue check verification during the open window of November 9-November 11, 2022. I retrieve 4.9m tweets from a sample of politically contentious accounts. I then compare engagement on contentious user posts before and after the Musk acquisition. I find that the period following the Musk acquisition saw a substantive increase in post engagement. There is no additional increase following blue tick verification. I explain the findings with reference to an increase in activity by a newly sympathetic user base.

cs.SI↗

arXiv2022

Explaining Recruitment to Extremism: A Bayesian Case-Control Approach

Who joins extremist movements? Answering this question poses considerable methodological challenges. Survey techniques are infeasible and selective samples provide no counterfactual. Recruits can be assigned to contextual units, but this is vulnerable to problems of ecological inference. In this article, we take inspiration from epidemiology and elaborate a technique that combines survey and ecological approaches. The multilevel Bayesian case-control design that we propose allows us to identify individual-level and contextual factors patterning the incidence of recruitment, while accounting for rare events, contamination, and spatial autocorrelation. We validate our approach by matching a sample of Islamic State (ISIS) fighters from nine MENA countries with representative population surveys enumerated shortly before recruits joined the movement. High status individuals in their early twenties with university education were more likely to join ISIS. There is more mixed evidence for relative deprivation. We provide software for applied researchers to implement our method.

stat.ME↗