Digitale Märkte und Öffentlichkeiten auf Plattformen

Dauerhafte URI für die Sammlung

https://www.weizenbaum-library.de/handle/id/309

Listen

Neueste Veröffentlichungen

Gerade angezeigt 1 - 5 von 71
  • Vorschaubild
    Item
    Can't LLMs Do That? Supporting Third-Party Audits Under the DSA: Exploring Large Language Models for Systemic Risk Evaluation of the Digital Services Act in an Interdisciplinary Setting
    (Association for Computing Machinery, 2025) Sekwenz, Marie-Therese; Gsenger, Rita; Stocker, Volker; Görnemann, Esther; Talypova, Dinara; Parkin, Simon; Greminger, Lea; Smaragdakis, Georgios
    This paper investigates the feasibility and potential role of using Large Language Models (LLMs) to support systemic risk audits under the European Union’s Digital Services Act (DSA). It examines how automated tools can enhance the work of DSA auditors and other ecosystem actors by enabling scalable, explainable, and legally grounded content analysis. An interdisciplinary expert workshop with twelve participants from legal, technical, and social science backgrounds explored prompting strategies for LLM-assisted auditing. Thematic analysis of the sessions identified key challenges and design considerations, including prompt engineering, model interpretability, legal alignment, and user empowerment. Findings highlight the potential of LLMs to improve annotation workflows and expand audit scale, while underscoring the continued importance of human oversight, iterative testing, and cross-disciplinary collaboration. This study offers practical insights for integrating AI tools into auditing processes and contributes to emerging methodologies for operationalizing systemic risk evaluations under the DSA.
  • Vorschaubild
    Item
    The Love of Large Numbers Revisited: A Coherence Model of the Popularity Bias
    (2020) Heck, Daniel W.; Seiling, Lukas; Bröder, Arndt
    Preferences are often based on social information such as experiences and recommendations of other people. The reliance on social information is especially relevant in the case of online shopping, where buying decisions for products may often be based on online reviews by other customers. Recently, Powell, Yu, DeWolf, and Holyoak (2017, Psychological Science, 28, 1432-1442) showed that, when deciding between two products, people do not consider the number of product reviews in a statistically appropriate way as predicted by a Bayesian model but rather exhibit a bias for popular products (i.e., products with many reviews). In the present work, we propose a coherence model of the cognitive mechanism underlying this empirical phenomenon. The new model assumes that people strive for a coherent representation of the available information (i.e., the average review score and the number of reviews). To test this theoretical account, we reanalyzed the data of Powell and colleagues and ran an online study with 244 participants using a wider range of stimulus material than in the original study. Besides replicating the popularity bias, the study provided clear evidence for the predicted coherence effect, that is, decisions became more confident and faster when the available information about popularity and quality was congruent.
  • Vorschaubild
    Item
    Radicalization within a network of misogynist extremists: a case study of an incel forum
    (2025) Coufal, Linda; Wedel, Lion
    Incels (involuntary celibates) are a group of people, linked to online misogyny and violent acts of terrorism, who mobilize around their inability to form romantic and/or sexual relationships. They have been shown to display signs of a violent extremist ideology. We conceptualize the ideology promoted by incels as misogynist and by bringing together different theories of gender and the gender order to formulate how the hetero-patriarchal and cisgenderist understanding of gender becomes an extremist worldview. We call this gender-based extremism misogynist extremism because misogyny is the most obviously violent structure of hetero-patriarchal gender order. Then, drawing on radicalization research and the social network analysis paradigm, we answer the research question: what are the communication patterns (network connections and actor attributes) that predict misogynist extremism among incels? We conduct our analysis on publicly visible posts from the forum incels.is, creating an undirected, unweighted network and then answering our research question using the auto-logistic actor attribute model to understand what individual attributes and network configurations predict user extremism. This study finds that extremists online form closed all-extremist communication triads. Consequently, they are significantly less likely to start new threads in the forum, suggesting that bonding social capital plays a more important role in an individual user’s extremism than bridging social capital.
  • Vorschaubild
    Item
    LGDE: Local Graph-based Dictionary Expansion
    (2025) Schindler, Juni; Jha, Sneha; Zhang, Xixuan; Buehling, Kilian; Heft, Annett; Barahona, Mauricio
    We present Local Graph-based Dictionary Expansion (LGDE), a method for data-driven discovery of the semantic neighbourhood of words using tools from manifold learning and network science. At the heart of LGDE lies the creation of a word similarity graph from the geometry of word embeddings followed by local community detection based on graph diffusion. The diffusion in the local graph manifold allows the exploration of the complex nonlinear geometry of word embeddings to capture word similarities based on paths of semantic association, over and above direct pairwise similarities. Exploiting such semantic neighbourhoods enables the expansion of dictionaries of pre-selected keywords, an important step for tasks in information retrieval, such as database queries and online data collection. We validate LGDE on two user-generated English-language corpora and show that LGDE enriches the list of keywords with improved performance relative to methods based on direct word similarities or co-occurrences. We further demonstrate our method through a real-world use case from communication science, where LGDE is evaluated quantitatively on the expansion of a conspiracy-related dictionary from online data collected and analysed by domain experts. Our empirical results and expert user assessment indicate that LGDE expands the seed dictionary with more useful keywords due to the manifold-learning-based similarity network.
  • Vorschaubild
    Item
    Epistemic authority in the digital public sphere. An integrative conceptual framework and research agenda
    (2025) Bartsch, Anne; Neuberger, Christoph; Stark, Birgit; Karnowski, Veronika; Maurer, Marcus; Pentzold, Christian; Quandt, Thorsten; Quiring, Oliver; Schemer, Christian
    We develop an integrative conceptual framework and research agenda for studying epistemic authorities in the digital age. Consulting epistemic authorities (e.g., professional experts, well-informed laypeople, technologies) can be an efficient fast-track to knowledge. To fulfill this functional role, those who claim epistemic authority need to be both subjectively recognized (have a perceived advantage in knowledge) and objectively justified (have an actual advantage in knowledge). In a digital media context, new and unconventional knowledge sources have emerged that can fulfill the functional role of epistemic authorities. But false authorities that disseminate misinformation have emerged as well while other sources with important knowledge remain unrecognized. We further analyze the functional role of epistemic intermediaries that can mitigate such problematic developments by correcting false authorities and by providing endorsement for unrecognized authorities. We conclude with a research agenda to study functional forms of epistemic authorities and epistemic intermediaries in the digital public sphere.