Since political discourse often revolves around determining which social groups are deemed deserving of resources, political social media discussions often reference them. Thereby, “social group” is a broad concept referring to collections of individuals who are categorized based on diverse criteria or attributions. Thus, social group mentions in text are prime examples of complex and ambiguous social concepts that have only recently garnered attention in natural language processing research, particularly within the emerging field of perspectivism. Following this line of research, we propose that disagreements between annotators provide important insights into what constitutes a social group. By analyzing diverging annotation judgments, we develop a taxonomy of label variation that distinguishes between ambiguities in identifying social group mentions and genuine annotation errors. We further investigate whether specific types of social group mentions are related to distinct types of label variation. Our findings reveal that linguistic and interpretative ambiguities represent a majority of variation when it comes to subjective and ambiguous concepts, underscoring the complex and multifaceted nature of social group mentions. This observation aligns with the perspectivist view, which posits that annotators' diverse perspectives can enrich labeled data by capturing multiple valid interpretations rather than forcing a single “correct” ground truth. To explore this, we present Reddit-Social Group Mentions (Reddit-SGM), a novel, non-aggregated dataset for studying social group mentions in texts. By embracing the perspectivist approach, our dataset lays the groundwork for the automatic identification of social group mentions, enabling a nuanced representation of this inherently ambiguous concept.

LLMs encode, convey, and perpetuate stereotypes. Prior computational research focuses on a small set of semantic axes investigated in social psychology, and operates on word embeddings produced by language models, leaving open which other semantic axes carry stereotypical associations in LLMs and how LLMs internally represent such axes. We introduce STEREODISCO, a framework that adapts the semantic differential method (Osgood et al., 1957) to the systematic study of stereotypes in LLM internal representations. STEREODISCO constructs approx. 2,000 candidate semantic axes from WordNet antonym synsets, recovers each as a geometric axis in the LLM's activation space via probing, and identifies stereotypical axes via a statistical test over concept projections. As a case study, we apply STEREODISCO to social group stereotypes with LLAMA-3-8B-INSTRUCT and MISTRAL-7B-INSTRUCT. We find that the two LLMs agree with each other on social group ratings more than with humans, suggesting that LLM-encoded stereotype content diverges from that documented in social psychology. We also discover stereotypical axes not investigated in prior work – including humble vs. proud, narrow-minded vs. broad-minded, and cowardly vs. brave, which human annotators independently confirm.

The same dataset can be analysed in different justifiable ways to answer the same research question, potentially challenging the robustness of empirical science1-3. In this crowd initiative, we investigated the degree to which research findings in the social and behavioural sciences are contingent on analysts' choices. We examined a stratified random sample of 100 studies published between 2009 and 2018, in which, for one claim per study, at least five reanalysts independently reanalysed the original data. The statistical appropriateness of the reanalyses was assessed in peer evaluations, and the robustness indicators were inspected along a range of research characteristics and study designs. We found that 34% of the independent reanalyses yielded the same result (within a tolerance region of ±0.05 Cohen's d) as the original report; with a four times broader tolerance region, this indicator increased to 57%. Of the reanalyses conducted, 74% reached the same conclusion as the original investigation, 24% yielded no effects or inconclusive results and 2% reported the opposite effect. This exploratory study indicates that the common single-path analyses in social and behavioural research should not be simply assumed to be robust to alternative analyses4. Therefore, we recommend the development and use of practices to explore and communicate this neglected source of uncertainty.

Translation of science to a general public is increasingly important in modern academia. Yet, there is little knowledge on whether and why scientists do so. Here, we provide an account of a population of social science professors in Germany (N = 2,207). We ask whether and to what extent scientists appear in German printed media (N = 26,729) as a result of cumulative advantage, reputation, and gender. We link bibliometric data on professors' careers and data on their appearances in printed media through unique, principled crosswalks of different databases. Departing from the literature on inequality in science, we develop hypotheses on how cumulative advantage, reputation, and gender relate to professors' media appearances. Employing a series of longitudinal logistic and linear regression analyses we find support for the majority of our conjectures. Cumulative advantages are particularly positively related to newspaper appearances. Once a scientist has been mentioned in the media they seem 'short-listed' and this dynamic is more pronounced among men rather than women professors. Reputable professors are also more likely to be in printed news. And men have a higher frequency of newspaper appearances than women, which seems driven by men more likely to be top mediagenic professors. We discuss the implications of these results for media practice and science evaluation.

The rise of populism concerns many political scientists and practitioners, yet the detection of its underlying language remains fragmentary. This paper aims to provide a reliable, valid, and scalable approach to measure populist stances. For that purpose, we created an annotated dataset based on parliamentary speeches of the German Bundestag (2013 to 2021). Following the ideational definition of populism, we label moralizing references to the virtuous people or the corrupt elite as core dimensions of populist language. To identify, in addition, how the thin ideology of populism is thickened, we annotate how populist statements are attached to left-wing or right-wing host ideologies. We then train a transformer-based model (PopBERT) as a multilabel classifier to detect and quantify each dimension. A battery of validation checks reveals that the model has a strong predictive accuracy, provides high qualitative face validity, matches party rankings of expert surveys, and detects out-of-sample text snippets correctly. PopBERT enables dynamic analyses of how German-speaking politicians and parties use populist language as a strategic device. Furthermore, the annotator-level data may also be applied in cross-domain applications or to develop related classifiers.

Policy actors (PAs) like nongovernmental organizations, political parties or governmental institutions strategically communicate on social media to gain attention and thus influence the public agenda. We argue that networks of PAs engaged in the same issues (i.e., a PA’s peer network ) are crucial to attracting the interest of a broad audience. Drawing on network theory, we posit that (i) ideological homophily, and (ii) the centrality and embeddedness in a PA’s peer network increase the attention received from all Twitter (now X) users. We investigate these premises by analyzing the European migration discourse on Twitter (2014–2020). The results of our study reveal that the centrality of PAs in their peer networks and ideologically similar relations considerably increase attention from the whole Twittersphere. These findings provide strong evidence that a PA’s role in its organizational peer network on social media governs the attention received in the overall discourse.

Social groups are central to political discussions. However, detecting social groups in text often relies on pre-determined socio-demographic categories or supervised learning methods that require extensive hand-labeled datasets. In this paper, we propose a methodology designed to leverage the potential of Large Language Models (LLMs) for the identification and annotation of social groups in text. The experiments show that open LLMs like Llama-2-70B-Chat and Mixtral-8-7B can reliably be used to annotate social groups in a few-shot scenario without the need for supervised learning. The automatically obtained annotations largely match human annotations on random samples from the Reddit Politosphere, resulting in micro-F1 scores of 0.71 and 0.83, respectively.

In this paper, we investigate how protective effects of intergenerational closure correspond with conflict networks in school classes. Taking a multilevel ecological perspective, we also consider networks' socio-spatial conditions. In a first step, we use ERGMs to analyze the association between parental contact and students' friendship ties, i.e., intergenerational closure (IC). Then, we utilize spatial regressions to analyze direct and moderating effects of a school's neighborhood on conflicts in the 135 German class networks (N = 3143 student measurements). In accordance with Coleman's theoretical notions, we find consistent negative effects of IC on the (standardized) density of the conflict networks. Moreover, we show that IC's impact is particularly strong in neighborhoods with a relatively high concentration of minorities. The results are in line with our theoretical considerations on multilevel network ecologies and the selective pressure of IC against conflict ties. Practically, our results provide evidence that fostering connections among parents (e.g., by implementing opportunity structures for parents to meet) might help to prevent deviating behavior in schools, especially in neighborhoods with relatively large proportions of ethnic minorities.

Digitization led to an enormous increase in the availability of visual data. As images are an important aspect of human communication, decades of social science research have analysed images, yet in mostly manual fashion with limited scaling capacities. In this work, we outline how recent advances in computer vision enable automated image analysis, allowing researchers to further unlock the potential of digital behavioural data. We introduce the field of computational social science and conduct a literature review of early studies using image recognition. We also highlight important aspects to be considered, such as computational demands and biases of computer vision models. Furthermore, in a case study, we examine the online behaviour of U.S. Members of Congress during the early COVID-19 pandemic in 2020. In particular, we focus on sharing images showing face masks as they are a crucial aspect of health and safety measures during the pandemic. Using Instagram data and models for detecting face masks, we find that temporal dynamics and party affiliation play a substantial role in the likelihood of sharing images of people wearing face masks: images with masks are more often posted after the introduction of mask mandates and Democratic party members are more likely to share images with masks. In addition, we find somewhat weaker to no differences regarding the age and gender of politicians.

Machine learning (ML) techniques have become one of the most successful scientific tools and changed the everyday life of people around the globe (e.g., search engines). A vast amount of digital data sources on human behaviour has emerged due to the rise of the internet and opened the door for computer scientists to apply ML on social phenomena. In the social sciences, however, the adoption of ML has been less enthusiastic. To investigate the relation of traditional statistics and ML, this paper shows how ML might be used as regression analysis. For that purpose, we illustrate what a typical social science approach might look like and how using ML techniques could contribute additional insights when it comes to estimators (non-linearity) or the assessment of model fit (predictive power). In particular, we reveal how epistemological differences shape the potential usage of ML in the social sciences and discuss the methodological trade-off of applying ML compared to traditional statistics.

Is the pursuit of interdisciplinary or innovative research beneficial or detrimental for the impact of early career researchers? We focus on young scholars as they represent an understudied population who have yet to secure a place within academia. Which effects promise higher scientific recognition (i.e., citations) is therefore crucial for the high-stakes decisions young researchers face. To capture these effects, we introduce measurements for interdisciplinarity and novelty that can be applied to a researcher's career. In contrast to previous studies investigating research impact on the paper level, hence, our paper focuses on a career perspective (i.e., the level of authors). To consider different disciplinary cultures, we utilize a comprehensive dataset on U.S. physicists (n = 4003) and psychologists (n = 4097), who graduated between 2008 and 2012, and traced their publication records. Our results indicate that conducting interdisciplinary research as an early career researcher in physics is beneficial, while it is negatively associated with research impact in psychology. In both fields, physics and psychology, early career researchers focusing on novel combinations of existing knowledge are associated with higher future impact. Taking some risks by deviating to a certain degree from mainstream paradigms seems therefore like a rewarding strategy for young scholars.

The European Union’s common public sphere project dates back to the 1960s and relies on Europeanisation through the gradual eradication of communication boundaries between its member countries. However, it is evident by now that Europeanisation of national public spheres is hard to achieve by increasing overlaps between national public spheres, synchronisation of news reporting across national boundaries, or diffusion of Europeanist norms into national politics. The European Union’s common public sphere project may hence be in danger. This calls for explorations of other imaginable models of the public sphere for Europe. Are there traces of other modes of transnational public sphere emerging in Europe? In this article, we explore a models of the transnational public sphere which is based on an alternative concept of Europeanisation derived from the cleavage theory. By drawing on social media data and employing tools of social network analysis, we demonstrate the empirical possibility of a cleavage model of the European public sphere.

Democratic politics builds on both clear differences and shared common ground. While the rise of digital media may have enabled more differences to be articulated, common ground is often seen as threatened by fragmentation of political debate, which some see as driven by news media. The relative importance of political actors (parties and politicians) in driving fragmentation has received less attention. In this paper, we compare how news media and political actors contribute to the fragmentation of online political debate on the basis of analysis of almost half a million election-related tweets collected during the 2017 French, German, and U.K. national elections. We employ a structural topic model to reduce online political debate to networks of topic overlap. Across the three countries with different political and media systems, we find news media are by far the most important actors in terms of creating and maintaining a common space of online political debate on Twitter. Our results also show that political actors, with some variation from country to country, contribute more to fragmentation as they focus on different topics while articulating clear differences. These findings underline the importance of complementing structural analysis of the rise of digital and social media with analysis of how important elite actors like news media and political parties/candidates use these media in different ways. Overall, we show how at least on Twitter, across three different countries with different media systems and political systems, news media create connection that contributes to commonality while political actors lay out clear differences that drive fragmentation.

Media discourse is often seen as an important condition of people's attitudes and perceptions. Despite a rich literature, however, it is not well understood how media exposure influences attitudes towards immigrants. In contrast to previous studies, we argue that people rely on 'availability heuristics' shaped by mass media. From that point of view, it is the specific content of media discourse on immigration that affects people's concerns. We use 'Structural Topic Models' to classify media content of more than 24.000 articles of leading German newspapers from 2001 to 2016. Utilizing 'linear fixed effect models' allows us to relate a person's concern towards immigration as reported in the German Socioeconomic Panel to prevalent topics discussed in print media while controlling for several confounding factors (e.g., party preferences, interest in politics, etc.). We find a robust relationship between topic salience and attitudes towards integration. Our results also reveal that specific topics with negative contents (e.g., domestic violence) to increase concerns, while others (e.g., scientific studies, soccer) decrease concerns substantially, underlining the importance of available information provided by media. In addition, people with higher education are generally less affected by media salience of topics.

Applications of machine learning (ML) in industry and natural sciences yielded some of the most impactful innovations of the last decade (for instance, artificial intelligence, gene prediction or search engines) and changed the everyday-life of many people. From a methodological perspective, we can differentiate between unsupervised machine learning (UML) and supervised machine learning (SML). While SML uses labeled data as input to train algorithms in order to predict outcomes of unlabeled data, UML detects underlying patterns in unlabeled observations by exploiting the statistical properties of the data. The possibilities of ML for analyzing large datasets are slowly finding their way into the social sciences; yet, it lacks systematic introductions into the epistemologically alien subject. I present applications of some of the most common methods for SML (i.e., logistic regression) and UML (i.e., topic models). A practical example offers social scientists a “how-to” description for utilizing both. With regard to SML, the case is made by predicting gender of a large dataset of sociologists. The proposed approach is based on open-source data and outperforms a popular commercial application (genderize.io). Utilizing the predicted gender in topic models reveals the stark thematic differences between male and female scholars that have been widely overlooked in the literature. By applying ML, hence, the empirical results shed new light on the longstanding question of gender-specific biases in academia.

AbstractIn this chapter, Raphael Heiberger and Michael Windzio examine which topics are important for major education international organization (IOs). IOs in the field of education follow different ideological paradigms in the global education discourse. Yet, it is an open question as to whether different types of IOs focus on different topics and thereby support different paradigms of education. Based on more than 1000 documents with over 40 million words published by the World Bank, UNESCO, the ILO, the OECD, ISESCO, and SEAMEO, they explore education issues addressed in this sample. Using standardized methods of quantitative text analysis and topic modeling, Heiberger and Windzio reveal that major topics found in these documents do indeed differ between the different types of organizations.

We investigate how sociology students garner recognition from niche field audiences through specialization. Our dataset comprises over 80,000 sociology-related dissertations completed at U.S. universities, as well as data on graduates’ pursuant publications. We analyze different facets of how students specialize—topic choice, focus, novelty, and consistency. To measure specialization types within a consistent methodological frame, we utilize structural topic modeling. These measures capture specialization strategies used at an early career stage. We connect them to a crucial long-term outcome in academia: becoming an advisor. Event-history models reveal that specific topic choices and novel combinations exhibit a positive influence, whereas focused theses make no substantial difference. In particular, theses related to the cultural turn, methods, or race are tied to academic careers that lead to mentorship. Thematic consistency of students’ publication track also has a strong positive effect on the chances of becoming an advisor. Yet, there are diminishing returns to consistency for highly productive scholars, adding important nuance to the well-known imperative of publish or perish in academic careers.

We map the topic structure of psychology utilizing a sample of over 500,000 abstracts of research articles and conference proceedings spanning two decades (1995–2015). To do so, we apply structural topic models to examine three research questions: (i) What are the discipline’s most prevalent research topics? (ii) How did the scientific discourse in psychology change over the last decades, especially since the advent of neurosciences? (iii) And was this change carried by high impact (HI) or less prestigious journals? Our results reveal that topics related to natural sciences are trending, while their ’counterparts’ leaning to humanities are declining in popularity. Those trends are even more pronounced in the leading outlets of the field. Furthermore, our findings indicate a continued interest in methodological topics accompanied by the ascent of neurosciences and related methods and technologies (e.g. fMRI’s). At the same time, other established approaches (e.g. psychoanalysis) become less popular and indicate a relative decline of topics related to the social sciences and the humanities.