OSF

Recommender system in X inadvertently profiles ideological positions of users

Studies on recommendations in social media have mainly analyzed the quality of recommended items (e.g., their diversity or biases) and the impact of recommendation policies (e.g., in comparison with purely chronological policies). We use a data donation program, collecting more than 2.5 million friend recommendations made to 682 volunteers on X over a year, to study instead how real-world recommenders learn, represent and process political and social attributes of users inside the so-called black boxes of AI systems. Using publicly available knowledge on the architecture of the recommender, we inferred the positions of recommended users in its embedding space. Leveraging ideology scaling calibrated with political survey data, we analyzed the political position of users in our study (N=26,509 among volunteers and recommended contacts) among several attributes, including age and gender. Our results show that the platform's recommender system produces a spatial ordering of users that is highly correlated with their Left-Right positions (Pearson rho=0.887, p-value < 0.0001), and that cannot be explained by socio-demographic attributes. These results open new possibilities for studying the interaction between human and AI systems. They also raise important questions linked to the legal definition of algorithmic profiling in data privacy regulation by blurring the line between active and passive profiling. We explore new constrained recommendation methods enabled by our results, limiting the political information in the recommender as a potential tool for privacy compliance capable of preserving recommendation relevance.

LLMs Can Infer Political Alignment from Online Conversations

Due to the correlational structure in our traits such as identities, cultures, and political attitudes, seemingly innocuous preferences such as following a band or using a specific slang, can reveal private traits. This possibility, especially when combined with massive, public social data and advanced computational methods, poses a fundamental privacy risk. Given our increasing data exposure online and the rapid advancement of AI are increasing the misuse potential of such risk, it is therefore critical to understand capacity of large language models (LLMs) to exploit it. Here, using online discussions on Debate.org and Reddit, we show that LLMs can reliably infer hidden political alignment, significantly outperforming traditional machine learning models. Prediction accuracy further improves as we aggregate multiple text-level inferences into a user-level prediction, and as we use more politics-adjacent domains. We demonstrate that LLMs leverage the words that can be highly predictive of political alignment while not being explicitly political. Our findings underscore the capacity and risks of LLMs for exploiting socio-cultural correlates.

Connective action and digital repression during China’s COVID-19 protests: a computational analysis of multilingual coordinated activity on Twitter - EPJ Data Science

In authoritarian contexts, social media serve as critical platforms for coordinating both protest and repression. This study centers on the unprecedented COVID-19 protests in the People’s Republic of China, which were extensively tweeted and suppressed through contesting narratives. We explore prominent themes, temporal dynamics, and linguistic patterns of coordinated communication during these events. Using a coordination detection algorithm, we identified 13,557 Twitter accounts involved in 739,819 instances of coordinated sharing during the protests. We then applied topic modeling to categorize the coordinated tweets into topics supporting either the protests or repression. Drawing on the theory of authoritarian publics, we classified protest-supporting topics into three categories: leadership-critical, policy-critical, and descriptive. Similarly, building on the digital repression typology, we categorized repression-supporting topics into government propaganda, distracting information, and demoralizing content. Within protest-supporting content, policy-critical tweets were the most widely shared across three analyzed languages. Leadership-critical tweets were more prominent in traditional Chinese, while descriptive tweets were more common in simplified Chinese. Repression-supporting content was most prevalent in English, followed by simplified Chinese, with demoralizing and distracting information dominating discourse. Government propaganda was the least frequent and appeared primarily in simplified Chinese. Community detection revealed that 85.4% of coordinated tweets were amplified by ten major communities, each organized around a single language and goal—either supporting protests or promoting repression. By combining multiple computational approaches, this study offers a comprehensive framework for content-centered analysis of online protest-repression dynamics and contributes to our understanding of connective action and digital repression in authoritarian contexts.

Google Scholar

Google Scholar