arXiv subjects

Emőke-Ágnes Horvát

Publications and source records attributed to Emőke-Ágnes Horvát.

At least 19 recordsLinked to original sources

arXiv2026

The Persistence of Retracted Papers on Wikipedia

Wikipedia serves as a key infrastructure for public access to scientific knowledge, but it faces challenges in maintaining the credibility of cited sources--especially when scientific papers are retracted. This paper investigates how citations to retracted research are handled on English Wikipedia. We construct a novel dataset that integrates Wikipedia revision histories with metadata from Retraction Watch, Crossref, Altmetric, and OpenAlex, identifying 1,181 citations of retracted papers. We find that 71.6% of the citations were initially problematic and in need of reader-facing repair, defined as those added before the paper's retraction (51.5%) or introduced afterwards without proper warning (20.1%). While many are eventually corrected, our analysis reveals that these citations persist for a median of 3.68 years (1,344 days). Through survival analysis, we find that bot-mediated flagging (RetractionBot), open access availability, pre-existing online visibility (e.g., Twitter/X mention counts), and page-level organization (e.g., number of categories on a Wikipedia page) are associated with a higher hazard of correction. Conversely, a paper's established scholarly authority--a higher academic citation count--is associated with a slower time to correction. Our findings highlight how the Wikipedia community supports collaborative maintenance but leaves gaps in citation-level repair. We contribute to CSCW research by advancing our understanding of this sociotechnical vulnerability, which takes the form of a community coordination challenge, and by offering design directions to support citation credibility at scale.

cs.HC↗

arXiv2025

Low-dimensional embeddings of high-dimensional data

Large collections of high-dimensional data have become nearly ubiquitous across many academic fields and application domains, ranging from biology to the humanities. Since working directly with high-dimensional data poses challenges, the demand for algorithms that create low-dimensional representations, or embeddings, for data visualization, exploration, and analysis is now greater than ever. In recent years, numerous embedding algorithms have been developed, and their usage has become widespread in research and industry. This surge of interest has resulted in a large and fragmented research field that faces technical challenges alongside fundamental debates, and it has left practitioners without clear guidance on how to effectively employ existing methods. Aiming to increase coherence and facilitate future work, in this review we provide a detailed and critical overview of recent developments, derive a list of best practices for creating and using low-dimensional embeddings, evaluate popular approaches on a variety of datasets, and discuss the remaining challenges and open problems in the field.

cs.LG↗

arXiv2025

Delving into LLM-assisted writing in biomedical publications through excess vocabulary

Large language models (LLMs) like ChatGPT can generate and revise text with human-level performance. These models come with clear limitations: they can produce inaccurate information, reinforce existing biases, and be easily misused. Yet, many scientists use them for their scholarly writing. But how wide-spread is such LLM usage in the academic literature? To answer this question for the field of biomedical research, we present an unbiased, large-scale approach: we study vocabulary changes in over 15 million biomedical abstracts from 2010--2024 indexed by PubMed, and show how the appearance of LLMs led to an abrupt increase in the frequency of certain style words. This excess word analysis suggests that at least 13.5% of 2024 abstracts were processed with LLMs. This lower bound differed across disciplines, countries, and journals, reaching 40% for some subcorpora. We show that LLMs have had an unprecedented impact on scientific writing in biomedical research, surpassing the effect of major world events such as the Covid pandemic.

cs.CL↗

arXiv2024

Collective dynamics behind success

Understanding the collective dynamics behind the success of ideas, products, behaviors, and social actors is critical for decision-making across diverse contexts, including hiring, funding, career choices, and the design of interventions for social change. Methodological advances and the increasing availability of big data now allow for a broader and deeper understanding of the key facets of success. Recent studies unveil regularities beneath the collective dynamics of success, pinpoint underlying mechanisms, and even enable predictions of success across diverse domains, including science, technology, business, and the arts. However, this research also uncovers troubling biases that challenge meritocratic views of success. This review synthesizes the growing, cross-disciplinary literature on the collective dynamics behind success and calls for further research on cultural influences, the origins of inequalities, the role of algorithms in perpetuating them, and experimental methods to further probe causal mechanisms behind success. Ultimately, these efforts may help to better align success with desired societal values.

physics.soc-ph↗

arXiv2024

The Unwanted Dissemination of Science: The Usage of Academic Articles as Ammunition in Contested Discursive Arenas on Twitter

Twitter is a common site of offensive language. Prior literature has shown that the emotional content of tweets can heavily impact their diffusion when discussing political topics. We extend prior work to look at offensive tweets that link to academic articles. Using a mixed methods approach, we identify three findings: firstly, offensive language is common in tweets that refer to academic articles, and vary widely by subject matter. Secondly, discourse analysis reveals that offensive tweets commonly use academic articles to promote or attack political ideologies. Lastly, we show that offensive tweets reach a smaller audience than their non-offensive counterparts. Our analysis of these offensive tweets reveal how academic articles are being shared on Twitter not for the sake of disseminating new knowledge, but rather to as argumentative tools in controversial and combative discourses.

cs.SI↗

arXiv2024

Online Engagement with Retracted Articles: Who, When, and How?

Retracted research discussed on social media can spread misinformation. Yet we lack an understanding of how retracted articles are mentioned by academic and non-academic users. This is especially relevant on Twitter due to the platform's prominent role in science communication. Here, we analyze the pre- and post-retraction differences in Twitter attention and engagement metrics for over 3,800 retracted English-language articles alongside comparable non-retracted articles. We subset these findings according to five user types detected by our supervised learning classifier: members of the public, academics, bots, science practitioners, and science communicators. We find that retracted articles receive greater user attention (tweet count) and engagement (likes, retweets, and replies) than non-retracted articles, especially among members of the public and bots, with the majority of user engagement happening before retraction. Our results highlight the prominent role of non-experts in discussions of retracted research and suggest an opportunity for social media platforms to contribute towards early detection of problematic scientific research online.

cs.SI↗

arXiv2024

Beyond Words: An Experimental Study of Signaling in Crowdfunding

Increasingly, crowdfunding is transforming financing for many people worldwide. Yet we know relatively little about how, why, and when funding outcomes are impacted by signaling between funders. We conduct two studies of N=500 and N=750 participants involved in crowdfunding to investigate the effect of certain characteristics of ``crowd signals'' on the decision to fund. We find that, under a variety of conditions, contributions of heterogeneous amounts arriving at varying time intervals are significantly more likely to be selected than homogeneous contribution amounts and times. The impact of signaling is strongest among participants who are susceptible to social influence. The effect is remarkably general across different project types, fundraising goals, participant interest in the projects, and participants' altruistic attitudes. Critically, the role of crowd signals in decision-making is typically unrecognized by participants. Our results underscore the fundamental nature of social signaling in crowdfunding, informing strategies for platforms, funders, and project creators.

cs.HC↗

arXiv2023

The Gender Gap in Scholarly Self-Promotion on Social Media

Self-promotion in science is ubiquitous but may not be exercised equally by men and women. Research on self-promotion in other domains suggests that, due to bias in self-assessment and adverse reactions to non-gender-conforming behaviors (``pushback''), women tend to self-promote less often than men. We test whether this pattern extends to scholars by examining self-promotion over six years using 23M Tweets about 2.8M research papers by 3.5M authors. Overall, women are about 28% less likely than men to self-promote their papers even after accounting for important confounds, and this gap has grown over time. Moreover, differential adoption of Twitter does not explain the gender gap, which is large even in relatively gender-balanced broad research areas, where bias in self-assessment and pushback are expected to be smaller. Further, the gap increases with higher performance and status, being most pronounced for productive women from top-ranked institutions who publish in high-impact journals. Critically, we find differential returns with respect to gender: while self-promotion is associated with increased tweets of papers, the increase is smaller for women than for men. Our findings suggest that self-promotion varies meaningfully by gender and help explain gender differences in the visibility of scientific ideas.

cs.DL↗

arXiv2023

Who benefits from altmetrics? The effect of team gender composition on the link between online visibility and citation impact

Online science dissemination has quickly become crucial in promoting scholars' work. Recent literature has demonstrated a lack of visibility for women's research, where women's articles receive fewer academic citations than men's. The informetric and scientometric community has briefly examined gender-based inequalities in online visibility. However, the link between online sharing of scientific work and citation impact for teams with different gender compositions remains understudied. Here we explore whether online visibility is helping women overcome the gender-based citation penalty. Our analyses cover the three broad research areas of Computer Science, Engineering, and Social Sciences, which have different gender representation, adoption of online science dissemination practices, and citation culture. We create a quasi-experimental setting by applying Coarsened Exact Matching, which enables us to isolate the effects of team gender composition and online visibility on the number of citations. We find that online visibility positively affects citations across research areas, while team gender composition interacts differently with visibility in these research areas. Our results provide essential insights into gendered citation patterns and online visibility, inviting informed discussions about decreasing the citation gap.

cs.CY↗

arXiv2023

Emergent Influence Networks in Good-Faith Online Discussions

Town hall-type debates are increasingly moving online, irrevocably transforming public discourse. Yet, we know relatively little about crucial social dynamics that determine which arguments are more likely to be successful. This study investigates the impact of one's position in the discussion network created via responses to others' arguments on one's persuasiveness in unfacilitated online debates. We propose a novel framework for measuring the impact of network position on persuasiveness, using a combination of social network analysis and machine learning. Complementing existing studies investigating the effect of linguistic aspects on persuasiveness, we show that the user's position in a discussion network influences their persuasiveness online. Moreover, the recognition of successful persuasion further increases this dominant network position. Our findings offer important insights into the complex social dynamics of online discourse and provide practical insights for organizations and individuals seeking to understand the interplay between influential positions in a discussion network and persuasive strategies in digital spaces.

cs.SI↗

arXiv2023

Understanding (Ir)rational Herding Online

Investigations of social influence in collective decision-making have become possible due to recent technologies and platforms that record interactions in far larger groups than could be studied before. Herding and its impact on decision-making are critical areas of practical interest and research study. However, despite theoretical work suggesting that it matters whether individuals choose who to imitate based on cues such as experience or whether they herd at random, there is little empirical analysis of this distinction. To demonstrate the distinction between what the literature calls "rational" and "irrational" herding, we use data on tens of thousands of loans from a well-established online peer-to-peer (p2p) lending platform. First, we employ an empirical measure of memory in complex systems to measure herding in lending. Then, we illustrate a network-based approach to visualize herding. Finally, we model the impact of herding on collective outcomes. Our study reveals that loan performance is not solely determined by whether the lenders engage in herding or not. Instead, the interplay between herding and the imitated lenders' prior success on the platform predicts loan outcomes. In short, herds led by expert lenders tend to pick loans that do not default. We discuss the implications of this under-explored aspect of herding for platform designers, borrowers, and lenders. Our study advances collective intelligence theories based on a case of high-stakes group decision-making online.

cs.SI↗

arXiv2023

Retracted Articles about COVID-19 Vaccines Enable Vaccine Misinformation on Twitter

Retracted scientific articles about COVID-19 vaccines have proliferated false claims about vaccination harms and discouraged vaccine acceptance. Our study analyzed the topical content of 4,876 English-language tweets about retracted COVID-19 vaccine research and found that 27.4% of tweets contained retraction-related misinformation. Misinformed tweets either ignored the retraction, or less commonly, politicized the retraction using conspiratorial rhetoric. To address this, Twitter and other social media platforms should expand their efforts to address retraction-related misinformation.

cs.SI↗

arXiv2023

Information Retention in the Multi-platform Sharing of Science

The public interest in accurate scientific communication, underscored by recent public health crises, highlights how content often loses critical pieces of information as it spreads online. However, multi-platform analyses of this phenomenon remain limited due to challenges in data collection. Collecting mentions of research tracked by Altmetric LLC, we examine information retention in the over 4 million online posts referencing 9,765 of the most-mentioned scientific articles across blog sites, Facebook, news sites, Twitter, and Wikipedia. To do so, we present a burst-based framework for examining online discussions about science over time and across different platforms. To measure information retention we develop a keyword-based computational measure comparing an online post to the scientific article's abstract. We evaluate our measure using ground truth data labeled by within field experts. We highlight three main findings: first, we find a strong tendency towards low levels of information retention, following a distinct trajectory of loss except when bursts of attention begin in social media. Second, platforms show significant differences in information retention. Third, sequences involving more platforms tend to be associated with higher information retention. These findings highlight a strong tendency towards information loss over time - posing a critical concern for researchers, policymakers, and citizens alike - but suggest that multi-platform discussions may improve information retention overall.

cs.CY↗

arXiv2023

Novelty and Cultural Evolution in Modern Popular Music

The ubiquity of digital music consumption has made it possible to extract information about modern music that allows us to perform large scale analysis of stylistic change over time. In order to uncover underlying patterns in cultural evolution, we examine the relationship between the established characteristics of different genres and styles, and the introduction of novel ideas that fuel this ongoing creative evolution. To understand how this dynamic plays out and shapes the cultural ecosystem, we compare musical artifacts to their contemporaries to identify novel artifacts, study the relationship between novelty and commercial success, and connect this to the changes in musical content that we can observe over time. Using Music Information Retrieval (MIR) data and lyrics from Billboard Hot 100 songs between 1974-2013, we calculate a novelty score for each song's aural attributes and lyrics. Comparing both scores to the popularity of the song following its release, we uncover key patterns in the relationship between novelty and audience reception. Additionally, we look at the link between novelty and the likelihood that a song was influential given where its MIR and lyrical features fit within the larger trends we observed.

cs.CY↗

arXiv2022

Dynamics of Cross-Platform Attention to Retracted Papers

Retracted papers often circulate widely on social media, digital news and other websites before their official retraction. The spread of potentially inaccurate or misleading results from retracted papers can harm the scientific community and the public. Here we quantify the amount and type of attention 3,851 retracted papers received over time in different online platforms. Comparing to a set of non-retracted control papers from the same journals, with similar publication year, number of co-authors and author impact, we show that retracted papers receive more attention after publication not only on social media, but also on heavily curated platforms, such as news outlets and knowledge repositories, amplifying the negative impact on the public. At the same time, we find that posts on Twitter tend to express more criticism about retracted than about control papers, suggesting that criticism-expressing tweets could contain factual information about problematic papers. Most importantly, around the time they are retracted, papers generate discussions that are primarily about the retraction incident rather than about research findings, showing that by this point papers have exhausted attention to their results and highlighting the limited effect of retractions. Our findings reveal the extent to which retracted papers are discussed on different online platforms and identify at scale audience criticism towards them. In this context, we show that retraction is not an effective tool to reduce online attention to problematic papers.

cs.CY↗

arXiv2021

Network Structures of Collective Intelligence: The Contingent Benefits of Group Discussion

Research on belief formation has produced contradictory findings on whether and when communication between group members will improve the accuracy of numeric estimates such as economic forecasts, medical diagnoses, and job candidate assessments. While some evidence suggests that carefully mediated processes such as the "Delphi method" produce more accurate beliefs than unstructured discussion, others argue that unstructured discussion outperforms mediated processes. Still others argue that independent individuals produce the most accurate beliefs. This paper shows how network theories of belief formation can resolve these inconsistencies, even when groups lack apparent structure as in informal conversation. Emergent network structures of influence interact with the pre-discussion belief distribution to moderate the effect of communication on belief formation. As a result, communication sometimes increases and sometimes decreases the accuracy of the average belief in a group. The effects differ for mediated processes and unstructured communication, such that the relative benefit of each communication format depends on both group dynamics as well as the statistical properties of pre-interaction beliefs. These results resolve contradictions in previous research and offer practical recommendations for teams and organizations.

econ.GN↗

arXiv2021

A Multi-Platform Study of Crowd Signals Associated with Successful Online Fundraising

The growing popularity of online fundraising (aka "crowdfunding") has attracted significant research on the subject. In contrast to previous studies that attempt to predict the success of crowdfunded projects based on specific characteristics of the projects and their creators, we present a more general approach that focuses on crowd dynamics and is robust to the particularities of different crowdfunding platforms. We rely on a multi-method analysis to investigate the correlates, predictive importance, and quasi-causal effects of features that describe crowd dynamics in determining the success of crowdfunded projects. By applying a multi-method analysis to a study of fundraising in three different online markets, we uncover general crowd dynamics that ultimately decide which projects will succeed. In all analyses and across the three different platforms, we consistently find that funders' behavioural signals (1) are significantly correlated with fundraising success; (2) approximate fundraising outcomes better than the characteristics of projects and their creators such as credit grade, company valuation, and subject domain; and (3) have significant quasi-causal effects on fundraising outcomes while controlling for potentially confounding project variables. By showing that universal features deduced from crowd behaviour are predictive of fundraising success on different crowdfunding platforms, our work provides design-relevant insights about novel types of collective decision-making online. This research inspires thus potential ways to leverage cues from the crowd and catalyses research into crowd-aware system design.

cs.HC↗

arXiv2021

Publishing patterns reflect political polarization in news media

Digital news outlets rely on a variety of outside contributors, from freelance journalists, to political commentators, to executives and politicians. These external dependencies create a network among news outlets, traced along the contributors they share. Using connections between outlets, we demonstrate how contributors' publishing trajectories tend to align with outlet political leanings. We also show how polarized clustering of outlets translates to differences in the topics of news covered and the style and tone of articles published. In addition, we demonstrate how contributors who cross partisan divides tend to focus on less explicitly political topics. This work addresses an important gap in the media polarization literature, by highlighting how structural factors on the production side of news media create an ecosystem shaped by political leanings, independent of the priorities of any one person or organization.

cs.SI↗