Data Science & Digitalisierung (Pröllochs)

Title: Emotions in online rumor diffusion

Authors: Nicolas Pröllochs, Dominik Bär (LMU Munich), Stefan Feuerriegel (LMU Munich)

Abstract:

Emotions are regarded as a dominant driver of human behavior, and yet their role in online rumor diffusion is largely unexplored. In this study, we empirically study the extent to which emotions explain the diffusion of online rumors. We analyze a large-scale sample of 107,014 online rumors from Twitter, as well as their cascades. For each rumor, the embedded emotions were measured based on eight so-called basic emotions from Plutchik’s wheel of emotions (i.e., anticipation–surprise, anger–fear, trust–disgust, joy–sadness). We then estimated using a generalized linear regression model how emotions are associated with the spread of online rumors in terms of (1) cascade size, (2) cascade lifetime, and (3) structural virality. Our results suggest that rumors conveying anticipation, anger, and trust generate more reshares, spread over longer time horizons, and become more viral. In contrast, a smaller size, lifetime, and virality is found for surprise, fear, and disgust. We further study how the presence of 24 dyadic emotional interactions (i.e., feelings composed of two emotions) is associated with diffusion dynamics. Here, we find that rumors cascades with high degrees of aggressiveness are larger in size, longer-lived, and more viral. Altogether, emotions embedded in online rumors are important determinants of the spreading dynamics.

Title: Community-Based Fact-Checking on Twitter’s Birdwatch Platform

Author: Nicolas Pröllochs

Abstract:

Misinformation undermines the credibility of social media and poses significant threats to modern societies. As a countermeasure, Twitter has recently introduced “Birdwatch,” a community-driven approach to address misinformation on Twitter. On Birdwatch, users can identify tweets they believe are misleading, write notes that provide context to the tweet and rate the quality of other users’ notes. In this work, we empirically analyze how users interact with this new feature. For this purpose, we collect all Birdwatch notes and ratings between the introduction of the feature in early 2021 and end of July 2021. We then map each Birdwatch note to the fact-checked tweet using Twitter’s historical API. In ad- addition, we use text mining methods to extract content characteristics from the text explanations in the Birdwatch notes (e. g., sentiment). Our empirical analysis yields the following main findings: (i) users more frequently file Birdwatch notes for misleading than not misleading tweets. These misleading tweets are primarily reported because of factual errors, lack of important context, or because they treat unverified claims as facts. (ii) Birdwatch notes are more helpful to other users if they link to trustworthy sources and if they embed a more positive sentiment. (iii) The social influence of the author of the source tweet is associated with differences in the level of user consensus. For influential users with many followers, Birdwatch notes yield a lower level of consensus among users and community-created fact checks are more likely to be seen as being incorrect and argumentative. Altogether, our findings can help social media platforms to formulate guidelines for users on how to write more helpful fact checks. At the same time, our analysis suggests that community-based fact-checking faces challenges regarding opinion speculation and polarization among the user base.

Hate Speech in the Political Discourse on Social Media: Disparities Across Parties, Gender, and Ethnicity
Proceedings of The Web Conference (WWW '22)

Paper 2: Solovev K, Pröllochs N (2022)
Moral Emotions Shape the Virality of COVID-19 Misinformation on Social Media
Proceedings of The Web Conference (WWW '22)

Preprints of both papers will become available soon.

Title: Are Longer Reviews Always More Helpful? Disentangling the Interplay Between Review Length and Argumentation Complexity

Authors: Bernhard Lutz (University of Freiburg), Nicolas Pröllochs, Dirk Neumann (University of Freiburg)

Abstract:

An overwhelming majority of previous works find longer product reviews to be more helpful than short reviews. In this paper, we build upon information overload theory and propose that longer reviews should not be assumed to be uniformly more helpful; instead, we argue that the effect depends on the complexity of the line of argumentation. To test this idea, we implement state-of-the-art machine learning methods that allow us to study the line of argumentation in reviews at the sentence-level. Our empirical analysis based on a dataset of Amazon customer reviews suggests that line of argumentation and review length are closely intertwined such that longer reviews with frequent changes between positive and negative arguments are perceived as less helpful. Our work has important implications for marketing professionals and retailer platforms that can utilize our results to optimize their customer feedback systems, enhance reviewer guidelines, and include more useful product reviews.

Im Rahmen der diesjährigen WiWi-Absolventenfeier wurde Prof. Dr. Nicolas Pröllochs mit dem Lehrpreis in der Kategorie Professor/Professorin ausgezeichnet. Dieser Preis wird von der Fachschaft Wirtschaftswissenschaften sowie des WiWi-Vereins für herausragende Lehrleistungen verliehen.

Die Professur dankt der Fachschaft, dem WiWi-Verein und den Studierenden für diese Auszeichung!

Online Emotions During the Storming of the US Capitol: Evidence from the Social Media Network Parler
ICWSM 2023 (preprint available via arXiv )

Abstract: The storming of the U.S. Capitol on January 6, 2021 has led to the killing of 5 people and is widely regarded as an attack on democracy. The storming was largely coordinated through social media networks such as Parler. Yet little is known regarding how users interacted on Parler during the storming of the Capitol. In this work, we examine the emotion dynamics on Parler during the storming with regard to heterogeneity across time and users. For this, we segment the user base into different groups (e.g., Trump supporters and QAnon supporters). We use affective computing to infer the emotions in the contents, thereby allowing us to provide a comprehensive assessment of online emotions. Our evaluation is based on a large-scale dataset from Parler, comprising of 717,300 posts from 144,003 users. We find that the user base responded to the storming of the Capitol with an overall negative sentiment. Akin to this, Trump supporters also expressed a negative sentiment and high levels of unbelief. In contrast to that, QAnon supporters did not express a more negative sentiment during the storming. We further provide a cross-platform analysis and compare the emotion dynamics on Parler and Twitter. Our findings point at a comparatively less negative response to the incidents on Parler compared to Twitter accompanied by higher levels of disapproval and outrage. Our contribution to research is three-fold: (1) We identify online emotions that were characteristic of the storming; (2) we assess emotion dynamics across different user groups on Parler; (3) we compare the emotion dynamics on Parler and Twitter. Thereby, our work offers important implications for actively managing online emotions to prevent similar incidents in the future.

Paper 2: Dominik Bär, Nicolas Pröllochs, Stefan Feuerriegel (2022)
Finding Qs: Profiling QAnon Supporters on Parler
ICWSM 2023 (preprint available via arXiv )

Abstract: The social media platform "Parler" has emerged into a prominent fringe community where a significant part of the user base are self-reported supporters of QAnon, a far-right conspiracy theory alleging that a cabal of elites controls global politics. QAnon is considered to have had an influential role in the public discourse during the 2020 U.S. presidential election. However, little is known about QAnon supporters on Parler and what sets them aside from other users. Building up on social identity theory, we aim at profiling the characteristics of QAnon supporters on Parler. We analyze a large-scale dataset with more than 600,000 profiles of English-speaking users on Parler. Based on users' profiles, posts, and comments, we then extract a comprehensive set of user features, linguistic features, network features, and content features. This allows us to perform user profiling and understand to what extent these features discriminate between QAnon and non-QAnon supporters on Parler. Our analysis is three-fold: (1) We quantify the number of QAnon supporters on Parler, finding that 34,913 users (5.5% of all users) openly report to support the conspiracy. (2) We examine differences between QAnon vs. non-QAnon supporters. We find that QAnon supporters differ statistically significantly from non-QAnon supporters across multiple dimensions. For example, they have, on average, a larger number of followers, followees, and posts, and thus have a large impact on the Parler network. (3) We use machine learning to identify which user characteristics discriminate QAnon from non-QAnon supporters. We find that user features, linguistic features, network features, and content features, can - to a large extent - discriminate QAnon vs. non-QAnon supporters on Parler. In particular, we find that user features are highly discriminatory, followed by content features and linguistic features.

Course: Text Mining (M. Sc.)

Title: Moralized language predicts hate speech on social media

Authors: Kirill Solovev & Nicolas Pröllochs

Abstract: Hate speech on social media threatens the mental health of its victims and poses severe safety risks to modern societies. Yet, the mechanisms underlying its proliferation, though critical, have remained largely unresolved. In this work, we hypothesize that moralized language predicts the proliferation of hate speech on social media. To test this hypothesis, we collected three datasets consisting of N = 691,234 social media posts and 35.5 million corresponding replies from Twitter that have been authored by societal leaders across three domains (politics, news media, and activism). Subsequently, we used textual analysis and machine learning to analyze whether moralized language carried in source tweets is linked to differences in the prevalence of hate speech in the corresponding replies. Across all three datasets, we consistently observed that higher frequencies of moral and moral-emotional words predict a higher likelihood of receiving hate speech. On average, each additional moral word was associated with between 10.66% and 16.48% higher odds of receiving hate speech. Likewise, each additional moral-emotional word increased the odds of receiving hate speech by between 9.35% and 20.63%. Furthermore, moralized language was a robust out-of-sample predictor of hate speech. These results shed new light on the antecedents of hate speech and may help to inform measures to curb its spread on social media.

The paper is available here (open access).

Title: Negativity drives online news consumption

Co-authors: Claire E. Robertson (NYU), Kaoru Schwarzenegger (ETH Zurich), Phillip Parnamets (Karolinska Institutet), Jay J. Van Bavel (NYU), Stefan Feuerriegel (LMU Munich)

Abstract: Online media is important for society in informing and shaping opinions, hence raising the question of what drives online news consumption. Here, we analyze the causal effect of negative and emotional words on news consumption using a large online dataset of viral news stories. Specifically, we conducted our analyses using a series of randomized controlled trials (N = 22,743). Our dataset comprises ∼105,000 different variations of news stories from Upworthy.com that generated ∼5.7 million clicks across more than 370 million overall impressions. Although positive words were slightly more prevalent than negative words, we found that negative words in news headlines increased consumption rates (and positive words decreased consumption rates). For a headline of average length, each additional negative word increased the click-through rate by 2.3% Our results contribute to a better understanding of why users engage with online media.

Paper available here .

Abstract:

Misinformation on social media presents a major threat to modern societies. While previous research has analyzed the virality across true and false social media posts, not every misleading post is necessarily equally viral. Rather, misinformation has different characteristics and varies in terms of its believability and harmfulness – which might influence its spread. In this study, we study how the perceived believability and harmfulness of misleading posts are associated with their virality on social media. Specifically, we empirically analyze a large sample of crowd-annotated social media posts from Twitter's Birdwatch platform, on which users can rate the believability and harmfulness of misleading tweets. To address our research questions, we implement an explanatory regression model and link the crowd ratings for believability and harmfulness to the virality of misleading posts on Twitter. Our findings imply that misinformation that is (i) easily believable and (ii) not particularly harmful is associated with more viral resharing cascades. These results offer insights into how different kinds of crowd fact-checked misinformation spreads and suggest that viral misinformation on social media is often not particularly concerning from the perspective of public safety. From a practical view, our findings may help platforms to develop more effective strategies to curb the proliferation of misleading posts on social media.

A preprint of the paper will become available soon.

The study has also been covered by Business Insider

• Media coverage in Psychology Today
• Research paper at PNAS Nexus (open access)

Title: New threats to society from free-speech social media platforms

Authors: Dominik Bär (LMU Munich), Nicolas Pröllochs, Stefan Feuerriegel (LMU Munich)

Abstract: In recent years, several free-speech social media platforms (so-called "alt-techs") have emerged, such as Parler, Gab, and Telegram. These platforms market themselves as alternatives to mainstream social media and proclaim "free-speech" due to the absence of content moderation, which has been attracting a large base of partisan users, extremists, and supporters of conspiracy theories. In this comment, we discuss some of the threats that emerge from such social media platforms and call for more policy efforts directed at understanding and countering the risks for society.

A preprint of the paper is available here .

• ARD
• FAZ
• Heise
• Deutschlandfunk
• ORF
• Psychology Today
• The Atlantic

Diffusion of Community Fact-Checked Misinformation on Twitter
Proceedings of the ACM on Human-Computer Interaction (CSCW), forthcoming

Abstract: The spread of misinformation on social media is a pressing societal problem that platforms, policymakers, and researchers continue to grapple with. As a countermeasure, recent works have proposed to employ non-expert fact-checkers in the crowd to fact-check social media content. While experimental studies suggest that crowds might be able to accurately assess the veracity of social media content, an understanding of how crowd fact-checked (mis-)information spreads is missing. In this work, we empirically analyze the spread of misleading vs. not misleading community fact-checked posts on social media. For this purpose, we employ a dataset of community-created fact-checks from Twitter's Birdwatch pilot and map them to resharing cascades on Twitter. Different from earlier studies analyzing the spread of misinformation listed on third-party fact-checking websites (e.g., Snopes), we find that community fact-checked misinformation is less viral. Specifically, misleading posts are estimated to receive 36.62% fewer retweets than not misleading posts. A partial explanation may lie in differences in the fact-checking targets: community fact-checkers tend to fact-check posts from influential user accounts with many followers, while expert fact-checks tend to target posts that are shared by less influential users. We further find that there are significant differences in virality across different sub-types of misinformation (e.g., factual errors, missing context, manipulated media). Moreover, we conduct a user study to assess the perceived reliability of (real-world) community-created fact-checks. Here, we find that users, to a large extent, agree with community-created fact-checks. Altogether, our findings offer insights into how misleading vs. not misleading posts spread and highlight the crucial role of sample selection when studying misinformation on social media.

Preprint available via arXiv

Paper 2: Nicolas Pröllochs, Stefan Feuerriegel (2023)
Mechanisms of True and False Rumor Sharing in Social Media: Collective Intelligence or Herd Behavior?
Proceedings of the ACM on Human-Computer Interaction (CSCW), forthcoming .

Abstract: Social media platforms disseminate extensive volumes of online content, including true and, in particular, false rumors. Previous literature has studied the diffusion of offline rumors, yet more research is needed to understand the diffusion of online rumors. In this paper, we examine the role of lifetime and crowd effects in social media sharing behavior for true vs. false rumors. Based on 126,301 Twitter cascades, we find that the sharing behavior is characterized by lifetime and crowd effects that explain differences in the spread of true as opposed to false rumors. All else equal, we find that a longer lifetime is associated with less sharing activities, yet the reduction in sharing is larger for false than for true rumors. Hence, lifetime is an important determinant explaining why false rumors die out. Furthermore, we find that the spread of false rumors is characterized by herding tendencies (rather than collective intelligence), whereby the spread of false rumors becomes proliferated at a larger retweet depth. These findings explain differences in the diffusion dynamics of true and false rumors and further offer practical implications for social media platforms.

Preprint available via arXiv

Title: Affective Information Processing of Fake News: Evidence From NeuroIS

Authors: Bernhard Lutz (University of Freiburg), Marc Adam (University of Newcastle), Stefan Feuerriegel (LMU Munich), Nicolas Pröllochs, Dirk Neumann (University of Freiburg)

Abstract: Fake news undermines individuals' ability to make informed decisions. However, the theoretical understanding of how users assess online news as real or fake has thus far remained incomplete. In particular, previous research cannot explain why users fall for fake news inadvertently and despite careful thinking. In this work, we study the role of affect when users assess online news as real or fake. We employ NeuroIS measurements as a complementary approach beyond self-reports, which allows us to capture affective responses in situ , i.e., directly in the moment they occur. We draw upon cognitive dissonance theory, which suggests that users experiencing affective responses avoid unpleasant information to reduce psychological discomfort. In our NeuroIS experiment, we measured affective responses based on electrocardiography and eye tracking. We find that lower heart rate variability and shorter mean fixation duration are associated with greater perceived fakeness and a higher probability of incorrect assessments, thus providing evidence of affective information processing. These findings imply that users may fall for fake news automatically and without even noticing. This has direct implications for information systems (IS) research and practice as effective countermeasures against fake news must account for affective information processing.

A preprint of the paper will become available soon.

Title: Russian propaganda on social media during the 2022 invasion of Ukraine

Authors: Dominique Geissler (LMU Munich), Dominik Bär (LMU Munich), Nicolas Pröllochs, Stefan Feuerriegel (LMU Munich)

Abstract:

The Russian invasion of Ukraine in February 2022 was accompanied by practices of information warfare, yet existing evidence is largely anecdotal while large-scale empirical evidence is lacking. Here, we analyze the spread of pro-Russian support on social media. For this, we collected N=349,455 messages from Twitter with pro-Russian support. Our findings suggest that pro-Russian messages received ~251,000 retweets and thereby reached around 14.4 million users. We further provide evidence that bots played a disproportionate role in the dissemination of pro-Russian messages and amplified its proliferation in early-stage diffusion. Countries that abstained from voting on the United Nations Resolution ES-11/1 such as India, South Africa, and Pakistan showed pronounced activity of bots. Overall, 20.28% of the spreaders are classified as bots, most of which were created at the beginning of the invasion. Together, our findings suggest the presence of a large-scale Russian propaganda campaign on social media and highlight the new threats to society that originate from it. Our results also suggest that curbing bots may be an effective strategy to mitigate such campaigns.

Preprint available via arXiv

Course: Text Mining (M. Sc.)

Media coverage (selection):

• FAZ
• Süddeutsche Zeitung
  
• ZDF
• Zeit
• Handelsblatt
• ... and many more

Our research paper published in EPJ Data Science is available here

Community Notes vs. Snoping: How the Crowd Selects Fact-Checking Targets on Social Media
ICWSM 2023 (preprint available via arXiv )

Abstract: Deploying links to fact-checking websites (so-called "snoping") is a common intervention that can be used by social media users to refute misleading claims. However, its real-world effect may be limited as it suffers from low visibility and distrust towards professional fact-checkers. As a remedy, Twitter launched its community-based fact-checking system Community Notes on which fact-checks are carried out by actual Twitter users and directly shown on the fact-checked tweets. Yet, an understanding of how fact-checking via Community Notes differs from snoping is absent. In this study, we analyze differences in how contributors to Community Notes and Snopers select their targets when fact-checking social media posts. For this purpose, we analyze two unique datasets from Twitter: (a) 25,912 community-created fact-checks from Twitter's Community Notes platform; and (b) 52,505 "snopes" that debunk tweets via fact-checking replies linking to professional fact-checking websites. We find that Notes contributors and Snopers focus on different targets when fact-checking social media content. For instance, Notes contributors tend to fact-check posts from larger accounts with higher social influence and are relatively less likely to endorse/emphasize the accuracy of not misleading posts. Fact-checking targets of Notes contributors and Snopers rarely overlap; however, those overlapping exhibit a high level of agreement in the fact-checking assessment. Moreover, we demonstrate that Snopers fact-check social media posts at a higher speed. Altogether, our findings imply that different fact-checking approaches -- carried out on the same social media platform -- can result in vastly different social media posts getting fact-checked. This has important implications for future research on misinformation, which should not rely on a single fact-checking approach when compiling misinformation datasets.

Title: Research can help to tackle AI-generated disinformation

Co-authors: Stefan Feuerriegel (LMU Munich), Renée DiResta (Stanford University), Josh A. Goldstein (Georgetown University), Srijan Kumar (Georgia Tech), Philipp Lorenz-Spreen (Max Planck Institute), Michael Tomz (Stanford University)

Abstract: Generative artificial intelligence (AI) tools have made it easy to create realistic disinformation that is hard to detect by humans and may undermine public trust. Some approaches used for assessing the reliability of online information may no longer work in the AI age. We offer suggestions for how research can help tackle the threats of AI-generated disinformation.

Link to paper: https://www.nature.com/articles/s41562-023-01726-2

Paper 1: Lutz B, Marc A, Feuerriegel S, Pröllochs N, Neumann D (2024)
Which Linguistic Cues Make People Fall for Fake News? A Comparison of Cognitive and Affective Processing
Proceedings of the ACM on Human-Computer Interaction (CSCW), forthcoming.

Abstract: Fake news on social media has large, negative implications for society. However, little is known about what linguistic cues make people fall for fake news and, hence, how to design effective countermeasures for social media. In this study, we seek to understand which linguistic cues make people fall for fake news. Linguistic cues (e.g., adverbs, personal pronouns, positive emotion words, negative emotion words) are important characteristics of any text and also affect how people process real vs. fake news. Specifically, we compare the role of linguistic cues across both cognitive processing (related to careful thinking) and affective processing (related to unconscious automatic evaluations). To this end, we performed a within-subject experiment where we collected neurophysiological measurements of 42 subjects while these read a sample of 40 real and fake news articles. During our experiment, we measured cognitive processing through eye fixations, and affective processing in situ through heart rate variability. We find that users engage more in cognitive processing for longer fake news articles, while affective processing is more pronounced for fake news written in analytic words. To the best of our knowledge, this is the first work studying the role of linguistic cues in fake news processing. Altogether, our findings have important implications for designing online platforms that encourage users to engage in careful thinking and thus prevent them from falling for fake news.

Preprint available via arXiv

Paper 2: Maarouf A, Pröllochs N, Feuerriegel S (2024)
The Virality of Hate Speech on Social Media
Proceedings of the ACM on Human-Computer Interaction (CSCW), forthcoming.

Abstract: Online hate speech is responsible for violent attacks such as, e.g., the Pittsburgh synagogue shooting in 2018, thereby posing a significant threat to vulnerable groups and society in general. However, little is known about what makes hate speech on social media go viral. In this paper, we collect N = 25,219 cascades with 65,946 retweets from X (formerly known as Twitter) and classify them as hateful vs. normal. Using a generalized linear regression, we then estimate differences in the spread of hateful vs. normal content based on author and content variables. We thereby identify important determinants that explain differences in the spreading of hateful vs. normal content. For example, hateful content authored by verified users is disproportionally more likely to go viral than hateful content from non-verified ones: hateful content from a verified user (as opposed to normal content) has a 3.5 times larger cascade size, a 3.2 times longer cascade lifetime, and a 1.2 times larger structural virality. Altogether, we offer novel insights into the virality of hate speech on social media.

Preprint available via arXiv

Abstract:

The Digital Services Act (DSA) requires large social media platforms in the EU to provide clear and specific information whenever they remove or restrict access to certain content. These "Statements of Reasons" (SoRs) are collected in the DSA Transparency Database to ensure transparency and scrutiny of content moderation decisions of the providers of online platforms. In this work, we empirically analyze 156 million SoRs within an observation period of two months to provide an early look at content moderation decisions of social media platforms in the EU. Our empirical analysis yields the following main findings: (i) There are vast differences in the frequency of content moderation across platforms. For instance, TikTok performs more than 350 times more content moderation decisions per user than X/Twitter. (ii) Content moderation is most commonly applied for text and videos, whereas images and other content formats undergo moderation less frequently. (ii) The primary reasons for moderation include content falling outside the platform's scope of service, illegal/harmful speech, and pornography/sexualized content, with moderation of misinformation being relatively uncommon. (iii) The majority of rule-breaking content is detected and decided upon via automated means rather than manual intervention. However, X/Twitter reports that it relies solely on non-automated methods. (iv) There is significant variation in the content moderation actions taken across platforms. Altogether, our study implies inconsistencies in how social media platforms implement their obligations under the DSA -- resulting in a fragmented outcome that the DSA is meant to avoid. Our findings have important implications for regulators to clarify existing guidelines or lay out more specific rules that ensure common standards on how social media providers handle rule-breaking content on their platforms.

A preprint of the paper is available on arXiv ( https://arxiv.org/abs/2312.04431 )

Title: Did the Roll-Out of Community Notes Reduce Engagement With Misinformation on X/Twitter?

Co-authors: Chuai Y, Tian H, Pröllochs N, Lenzini G (University of Luxembourg)

Abstract: Developing interventions that successfully reduce engagement with misinformation on social media is challenging. One intervention that has recently gained great attention is X/Twitter's Community Notes (previously known as "Birdwatch"). Community Notes is a crowdsourced fact-checking approach that allows users to write textual notes to inform others about potentially misleading posts on X/Twitter. Yet, empirical evidence regarding its effectiveness in reducing engagement with misinformation on social media is missing. In this paper, we perform a large-scale empirical study to analyze whether the introduction of the Community Notes feature and its roll-out to users in the US and around the world have reduced engagement with misinformation on X/Twitter in terms of retweet volume and likes. We employ Difference-in-Differences (DiD) models and Regression Discontinuity Design (RDD) to analyze a comprehensive dataset consisting of all fact-checking notes and corresponding source tweets since the launch of Community Notes in early 2021. Although we observe a significant increase in the volume of fact-checks carried out via Community Notes, particularly for tweets from verified users with many followers, we find no evidence that the introduction of Community Notes significantly reduced engagement with misleading tweets on X/Twitter. Rather, our findings suggest that Community Notes might be too slow to effectively reduce engagement with misinformation in the early (and most viral) stage of diffusion. Our work emphasizes the importance of evaluating fact-checking interventions in the field and offers important implications to enhance crowdsourced fact-checking strategies on social media.

Preprint available via arXiv

Two papers have been accepted for poster presentation at the ACM Web Science Conference (WebSci 24). Chiara Drolsbach will present the result from a large-scale survey experiment analyzing users' trust in different types fact-checking interventions (e.g., community notes, expert fact-checks). Kirill Solovev will present his empirical work analyzing drivers of hate speech on social work.

Title: Community notes increase trust in fact-checking on social media

Authors: Chiara Drolsbach, Kirill Solovev & Nicolas Pröllochs

Abstract: Community-based fact-checking is a promising approach to fact-check social media content at scale. However, an understanding of whether users trust community fact-checks is missing. Here, we presented n = 1810 Americans with 36 misleading and non-misleading social media posts and assessed their trust in different types of fact-checking interventions. Participants were randomly assigned to treatments where misleading content was either accompanied by simple (i.e., context-free) misinformation flags in different formats (expert flags or community flags), or by textual "community notes" explaining why the fact-checked post was misleading. Across both sides of the political spectrum, community notes were perceived as significantly more trustworthy than simple misinformation flags. Our results further suggest that the higher trustworthiness primarily stemmed from the context provided in community notes (i.e., fact-checking explanations) rather than generally higher trust towards community fact-checkers. Community notes also improved the identification of misleading posts. In sum, our work implies that context matters in fact-checking and that community notes might be an effective approach to mitigate trust issues with simple misinformation flags.

The full paper is available here (open access).

Our paper "Community notes increase trust in fact-checking on social media" (co-authored by Chiara Drolsbach, Kirill Solovev & Nicolas Pröllochs) has won the "Best Poster Award" at ACM WebSci '24 in Stuttgart, Germany. At WebSci '24, Chiara Drolsbach presented the results from a preregistered survey experiment analyzing users' trust in different types fact-checking interventions on social media (e.g., community notes, expert fact-checks). The findings imply that community notes might be an effective approach to mitigate trust issues with simple misinformation flags.

The full paper (preprint) is available via OSF

Interview mit der FAZ: https://www.faz.net/aktuell/wissen/leben-gene/social-media-sind-fake-news-viraler-als-die-wahrheit-19871862.html

Unsere Studie in PNAS Nexus: https://academic.oup.com/pnasnexus/article-pdf/3/7/pgae217/58364142/pgae217.pdf

Course: Text Mining (M. Sc.)

Title: A Fused Large Language Model for Predicting Startup Success

Co-authors: Abdurahman Maarouf (LMU Munich), Stefan Feuerriegel (LMU Munich)

Abstract: Investors are continuously seeking profitable investment opportunities in startups and, hence, for effective decision-making, need to predict a startup’s probability of success. Nowadays, investors can use not only various fundamental information about a startup (e.g., the age of the startup, the number of founders, and the business sector) but also textual description of a startup’s innovation and business model, which is widely available through online venture capital (VC) platforms such as Crunchbase. To support the decision making of investors, we develop a machine learning approach with the aim of locating successful startups on VC platforms. Specifically, we develop, train, and evaluate a tailored, fused large language model to predict startup success. Thereby, we assess to what extent self-descriptions on VC platforms are predictive of startup success. Using 20,172 online profiles from Crunchbase, we find that our fused large language model can predict startup success, with textual self-descriptions being responsible for a significant part of the predictive power. Our work provides a decision support tool for investors to find profitable investment opportunities.

The paper is available here (open access): https://doi.org/10.1016/j.ejor.2024.09.011

Einen Link zur Stellenanzeige finden Sie hier .

Die Bewerbungsfrist ist der 12.11.2024.

• Featured Paper 1: Community notes increase trust in fact-checking on social media ( https://academic.oup.com/pnasnexus/article/3/7/pgae217/7686087 )
• Featured Paper 2: Community notes reduce the spread of misleading posts on X ( https://osf.io/preprints/osf/3a4fe )

Is Fact-Checking Politically Neutral? Asymmetries in How U.S. Fact-Checking Organizations Pick Up False Statements Mentioning Political Elites
ICWSM 2025 (preprint available via arXiv )

Abstract: Political elites play an important role in the proliferation of online misinformation. However, an understanding of how fact-checking platforms pick up politicized misinformation for fact-checking is still in its infancy. Here, we conduct an empirical analysis of mentions of U.S. political elites within fact-checked statements. For this purpose, we collect a comprehensive dataset consisting of 35,014 true and false statements that have been fact-checked by two major fact-checking organizations (Snopes, PolitiFact) in the U.S. between 2008 and 2023, i.e., within an observation period of 15 years. Subsequently, we perform content analysis and explanatory regression modeling to analyze how veracity is linked to mentions of U.S. political elites in fact-checked statements. Our analysis yields the following main findings: (i) Fact-checked false statements are, on average, 20% more likely to mention political elites than true fact-checked statements. (ii) There is a partisan asymmetry such that fact-checked false statements are 88.1% more likely to mention Democrats, but 26.5% less likely to mention Republicans, compared to fact-checked true statements. (iii) Mentions of political elites in fact-checked false statements reach the highest level during the months preceding elections. (iv) Fact-checked false statements that mention political elites carry stronger other-condemning emotions and are more likely to be pro-Republican, compared to fact-checked true statements. In sum, our study offers new insights into understanding mentions of political elites in false statements on U.S. fact-checking platforms, and bridges important findings at the intersection between misinformation and politicization.

Title: Using natural language processing to analyse text data in behavioural science

Co-authors: Feuerriegel S (LMU Munich), Maarouf A (LMU Munich), Bär D (LMU Munich), Geissler D (LMU Munich), Schweisthal J (LMU Munich), Robertson C (NYU), Rathje S (NYU), Hartmann J (TUM), Mohammad S (NRC Canada), Netzer O (Columbia Business School), Siegel A (University of Colorado), Plank B (LMU Munich), Van Bavel J (NYU)

Abstract: Language is a uniquely human trait at the core of human interactions. The language people use often reflects their personality, intentions and state of mind. With the integration of the Internet and social media into everyday life, much of human communication is documented as written text. These online forms of communication (for example, blogs, reviews, social media posts and emails) provide a window into human behaviour and therefore present abundant research opportunities for behavioural science. In this Review, we describe how natural language processing (NLP) can be used to analyse text data in behavioural science.  First, we review applications of text data in behavioural science. Second, we describe the NLP pipeline and explain the underlying modelling approaches (for example, dictionary-based approaches and large language models). We discuss the advantages and disadvantages of these methods for behavioural science, in particular with respect to the trade-off between interpretability and accuracy. Finally, we provide actionable recommendations for using NLP to ensure rigour and reproducibility.

Link to paper: https://rdcu.be/d5oda

Community Fact-Checks Trigger Moral Outrage in Replies to Misleading Posts on Social Media
CHI '25 (preprint available via arXiv )

Abstract: Displaying community fact-checks is a promising approach to reduce engagement with misinformation on social media. However, how users respond to misleading content emotionally after community fact-checks are displayed on posts is unclear. Here, we employ quasi-experimental methods to causally analyze changes in sentiments and (moral) emotions in replies to misleading posts following the display of community fact-checks. Our evaluation is based on a large-scale panel dataset comprising N=2,225,260 replies across 1841 source posts from X's Community Notes platform. We find that informing users about falsehoods through community fact-checks significantly increases negativity (by 7.3%), anger (by 13.2%), disgust (by 4.7%), and moral outrage (by 16.0%) in the corresponding replies. These results indicate that users perceive spreading misinformation as a violation of social norms and that those who spread misinformation should expect negative reactions once their content is debunked. We derive important implications for the design of community-based fact-checking systems.

Course: Applied AI (M. Sc.)

Artificial Intelligence (AI) is transforming the businesses by unlocking new opportunities for efficiency and data-driven decision-making. The master’s course on “Applied AI” provides students with an overview of the field of AI with a focus on real-world applications. Students will learn the end-to-end process of preparing data, implementing machine learning models, and evaluating their performance. The course will provide hands-on coding examples, equipping students with the necessary skills to implement these techniques independently. At the end of the course, participants will be familiar with the most important concepts, principles, algorithms, and challenges in applied AI.

The main objectives of this course are to:

1. Understand the basic concepts of AI and machine learning and their relevance in business contexts
2. Obtain an overview of different methods, algorithms, and software tools for applied AI
3. Learn how to train and evaluate AI methods on real-world datasets
4. Understand limits and challenges associated with contemporary AI methods, including ethical considerations and biases

Organisation:

• Module codes: 02-BWL/VWL:MSc-B11-Extra2
• Lecturer: Prof. Dr. Nicolas Pröllochs (BWL XI)
• Course format: Lecture (6 CP)
• Term: Summer semester 25
• Language: English
• Grading: Presentation
• Schedule: See course flyer

The number of participants is limited. The application deadline is April 14, 2025. Details about the application process can be found in the course flyer .

Title: The role of social media ads for election outcomes: Evidence from the 2021 German election

Authors: Dominik Bär (LMU Munich), Nicolas Pröllochs, Stefan Feuerriegel (LMU Munich)

Abstract: Social media ads have become a key communication channel in politics. However, the relationship between political ads from social media and election outcomes is not fully understood. Here, we aim to estimate the association between online political advertising and election outcomes during the 2021 German federal election. For this, we analyze a large-scale dataset of 21,641 political ads from Facebook and Instagram that received ≈126 million impressions. Using regression analysis, we show that political advertising on social media has a positive relationship with a candidate's election outcome and may even sway elections. All else equal, ≈200,000 additional impressions are predicted to increase a candidate's votes by 2.1%. We further use a causal sensitivity analysis to evaluate how unobserved confounding may affect our estimates. We find that the estimated impact of ads cannot be reasonably explained away, highlighting the significance of social media for election outcomes.

The full paper is available here (open access).

Link zum Podcast: https://www.deutschlandfunk.de/kriege-krise-katastrophen-berichten-medien-zu-wenig-ueber-positives-100.html

• The Atlantic
• TIME Magazine

Interview im 1E9 Magazin: https://www.1e9.community/magazin/nutzen-oder-schaden-die-sozialen-medien-der-demokratie-prof-proellochs

Our study "Characterizing Deepfakes on X" has been accepted for presentation at the International Conference on Computational Social Science (IC2S2 '25). Chiara Drolsbach will present the result from a large-scale empirical analysis of AI-generated misinformation on the social media platform X.

Title: References to unbiased sources increase the helpfulness of community fact-checks

Authors: Kirill Solovev & Nicolas Pröllochs

Abstract: Community-based fact-checking is a promising approach to address misinformation on social media at scale. However, an understanding of what makes community-created fact-checks helpful to users is still in its infancy. In this paper, we analyze the determinants of the helpfulness of community-created fact-checks. For this purpose, we draw upon a unique dataset of real-world community-created fact-checks and helpfulness ratings from X’s (formerly Twitter) Community Notes platform. Our empirical analysis implies that the key determinant of helpfulness in community-based fact-checking is whether users provide links to external sources to underpin their assertions. On average, the odds for community-created fact-checks to be perceived as helpful are 2.33 times higher if they provide links to external sources. Furthermore, we demonstrate that the helpfulness of community-created fact-checks varies depending on their level of political bias. Here, we find that community-created fact-checks linking to high-bias sources (of either political side) are perceived as significantly less helpful. This suggests that the rating mechanism on the Community Notes platform successfully penalizes one-sidedness and politically motivated reasoning. These findings have important implications for social media platforms, which can utilize our results to optimize their community-based fact-checking systems.

A preprint of the paper is available via arXiv .

Course: Text Mining (M. Sc.)

Course: Applied AI (M. Sc.)

Artificial Intelligence (AI) is transforming the businesses by unlocking new opportunities for efficiency and data-driven decision-making. The master’s course on “Applied AI” provides students with an overview of the field of AI with a focus on real-world applications. Students will learn the end-to-end process of preparing data, implementing machine learning models, and evaluating their performance. The course will provide hands-on coding examples, equipping students with the necessary skills to implement these techniques independently. At the end of the course, participants will be familiar with the most important concepts, principles, algorithms, and challenges in applied AI.

The main objectives of this course are to:

1. Understand the basic concepts of AI and machine learning and their relevance in business contexts
2. Obtain an overview of different methods, algorithms, and software tools for applied AI
3. Learn how to train and evaluate AI methods on real-world datasets
4. Understand limits and challenges associated with contemporary AI methods, including ethical considerations and biases

Organisation:

• Module codes: 02-BWL/VWL:MSc-B11-Extra2
• Lecturer: Prof. Dr. Nicolas Pröllochs (BWL XI)
• Course format: Lecture (6 CP)
• Term: Winter semester 25/26
• Language: English
• Grading: Presentation
• Schedule: See course flyer

The number of participants is limited. The application deadline is October 2, 2025. Details about the application process can be found in the course flyer .

Einen Link zur Stellenanzeige finden Sie hier .

Die Bewerbungsfrist ist der 04.12.2025.

Paper 1: Bobek M, Pröllochs N (2026)
Community Fact-Checks Do Not Break Follower Loyalty
Proceedings of the ACM Web Conference (WWW), forthcoming.

Abstract: Major social media platforms increasingly adopt community-based fact-checking to address misinformation on their platforms. While previous research has largely focused on its effect on engagement (e.g., reposts, likes), an understanding of how fact-checking affects a user's follower base is missing. In this study, we employ quasi-experimental methods to causally assess whether users lose followers after their posts are corrected via community fact-checks. Based on time-series data on follower counts for N=3516 community fact-checked posts from X, we find that community fact-checks do not lead to meaningful declines in the follower counts of users who post misleading content. This suggests that followers of spreaders of misleading posts tend to remain loyal and do not view community fact-checks as a sufficient reason to disengage. Our findings underscore the need for complementary interventions to more effectively disincentivize the production of misinformation on social media.

Preprint available via arXiv

Paper 2: Chuai Y, Lenzini G, Pröllochs N (2026)
Consensus Stability of Community Notes on X
Proceedings of the ACM Web Conference (WWW), forthcoming.

Abstract: Community-based fact-checking systems, such as Community Notes on X (formerly Twitter), aim to mitigate online misinformation by surfacing annotations judged helpful by contributors with diverse viewpoints. While prior work has shown that the platform’s bridging-based algorithm effectively selects helpful notes at the time of display, little is known about how evaluations change after notes become visible. Using a large-scale dataset of 437,396 Community Notes and 35 million ratings from over 580,000 contributors, we examine the stability of helpful notes and the rating dynamics that follow their initial display. We find that 30.2% of displayed notes later lose their helpful status and disappear. Using interrupted time series models, we further show that note display triggers a sharp increase in rating volume and a significant shift in rating leaning, but these effects differ across rater groups. Contributors with viewpoints similar to note authors tend to increase supportive ratings, while dissimilar contributors increase negative ratings, producing systematic post-display polarization. Counterfactual analyses suggest that this post-display polarization, particularly from dissimilar raters, plays a substantial role in note disappearance. These findings highlight the vulnerability of consensus-based fact-checking systems to polarized rating behavior and suggest pathways for improving their resilience.

Title: Engagement with political videos on TikTok during the 2025 German federal election

Authors: Kirill Solovev, Chiara Drolsbach, Emma Demirel, Nicolas Pröllochs

Abstract:

Short-form video platforms like TikTok reshape how politicians communicate and have become important tools for electoral campaigning. Yet it remains unclear what kinds of political messages gain traction in these fast-paced, algorithmically curated environments, which are particularly popular among younger audiences. In this study, we use computational content analysis to analyze a comprehensive dataset of N= 25,292 TikTok videos posted by German politicians in the run-up to the 2025 German federal election. Our empirical analysis shows that videos expressing negative emotions (e.g., anger, disgust) and outgroup animosity were significantly more likely to generate engagement than those emphasizing positive emotion, relatability, or identity. Furthermore, ideologically extreme parties (on both sides of the political spectrum) were both more likely to post this type of content and more successful in generating engagement than centrist parties. Taken together, these findings suggest that divisive political communication tends to receive higher engagement than unifying messages on TikTok, thereby potentially benefiting extreme actors who are more inclined to capitalize on this logic.

Paper available here (open access)

Title: Community-based fact-checking reduces the spread of misleading posts on X (formerly Twitter)

Co-authors: Yuwei Chuai, Moritz Pilarski, Thomas Renault, David Restrepo-Amariles, Aurore Troussel-Clément, Gabriele Lenzini & Nicolas Pröllochs

Abstract: Community-based fact-checking is a promising approach to correct misleading posts at scale. Yet, causal evidence regarding its effectiveness in reducing the spread of misinformation on social media is missing. Here, we perform a large-scale empirical study to analyze whether community notes reduce the spread of misleading posts on X (formerly Twitter). Using a Difference-in-Differences design and repost time series data for N = 237,180 (community fact-checked) cascades that have been reposted more than 431 million times, we find that exposing users to community notes reduces the subsequent spread of misleading posts by, on average, 61.2%. The effect is pronounced across the board but significantly weaker for posts from influential accounts and political content. Additionally, community notes increase the odds that users delete their misleading posts by 94.3%. Although community notes are broadly effective in reducing the spread of posts once annotated, they often appear too late to intervene in the early (and most viral) stage of the diffusion. As a result, their system-wide effect is more modest, lowering total engagement with misleading posts by 14.9%. Our work provides important insights that can inform future initiatives aimed at increasing the effectiveness of community-based fact-checking approaches on social media.

Link to paper (open access): https://doi.org/10.1038/s41467-026-72597-0

Einen Link zur Stellenanzeige finden Sie hier .

Die Bewerbungsfrist ist der 15.07.2026.

Course: Text Mining (M. Sc.)

Course: Applied AI (M. Sc.)

Artificial Intelligence (AI) is transforming the businesses by unlocking new opportunities for efficiency and data-driven decision-making. The master’s course on “Applied AI” provides students with an overview of the field of AI with a focus on real-world applications. Students will learn the end-to-end process of preparing data, implementing machine learning models, and evaluating their performance. The course will provide hands-on coding examples, equipping students with the necessary skills to implement these techniques independently. At the end of the course, participants will be familiar with the most important concepts, principles, algorithms, and challenges in applied AI.

The main objectives of this course are to:

1. Understand the basic concepts of AI and machine learning and their relevance in business contexts
2. Obtain an overview of different methods, algorithms, and software tools for applied AI
3. Learn how to train and evaluate AI methods on real-world datasets
4. Understand limits and challenges associated with contemporary AI methods, including ethical considerations and biases

Organisation:

• Module codes: 02-BWL/VWL:MSc-B11-Extra2
• Lecturer: Prof. Dr. Nicolas Pröllochs (BWL XI)
• Course format: Lecture (6 CP)
• Term: Winter semester 26/27
• Language: English
• Grading: Term project (incl. presentation)
• Schedule: See course flyer

The number of participants is limited. The application deadline is October 5, 2026. Details about the application process can be found in the course flyer .