No description available.

La diffusion des Fake News sur les réseaux sociaux est devenue un problème central ces dernières années. Notamment, hoaxy rapporte que les efforts de fact checking prennent généralement 10 à 20 heures pour répondre à une fake news, et qu'il y a un ordre de magnitude en plus de fake news que de fact checking. Le fact checking automatique pourrait aider en accélérant le travail humain et en surveillant les tendances dans les fake news. Dans un effort contre la désinformation, nous résumons le domaine de Fact Checking Automatique basé sur le contenu en 3 approches: les modèles avec aucune connaissances externes, les modèles avec un Graphe de Connaissance et les modèles avec une Base de Connaissance. Afin de rendre le Fact Checking Automatique plus accessible, nous présentons pour chaque approche une architecture efficace avec le poids en mémoire comme préoccupation, nous discutons aussi de comment chaque approche peut être appliquée pour faire usage au mieux de leur charactéristiques. Nous nous appuyons notamment sur la version distillée du modèle de langue BERT tinyBert, combiné avec un partage fort des poids sur 2 approches pour baisser l'usage mémoire en préservant la précision. / The spreading of fake news on social media has become a concern in recent years. Notably, hoaxy found that fact checking generally takes 10 to 20 hours to respond to a fake news, and that there is one order of magnitude more fake news than fact checking. Automatic fact checking could help by accelerating human work and monitoring trends in fake news. In the effort against disinformation, we summarize content-based automatic fact-checking into 3 approaches: models with no external knowledge, models with a Knowledge Graph and models with a Knowledge Base. In order to make Automatic Fact Checking more accessible, we present for each approach an effective architecture with memory footprint in mind and also discuss how they can be applied to make use of their different characteristics. We notably rely on distilled version of the BERT language model tinyBert, combined with hard parameter sharing on two approaches to lower memory usage while preserving the accuracy.

Social media platforms create rich social structures, expand users' boundaries of social networks and revolutionize traditional forms of communications, social interactions and social relationships. These platforms not only facilitate the creation and sharing of news and information, but they also drive various kinds of businesses models, processes and operations, knowledge sharing, marketing strategies for brand management and socio-political discourses essential for healthy and democratic functions. As such, social media has greater implications on organizations and society brought about by individuals' social media usage patterns, and therefore, calls for further investigations. The main objective of this dissertation is to explore and offer insights into such social media usage and information sharing behaviors via data driven examination of various theories. This dissertation involves three studies that focus on factors that explain individuals' three different social media usage behaviors. Essay 1 investigates individuals' perceived importance of online affiliation, self-esteem, self-regulation and risk-benefit structure as antecedents of users' geo-tagging behavior on social media. Essay 2 examines the role of online news quality, source credibility, individuals' perception towards online civic engagement, attitude towards news sharing and social influences to understand users' news sharing behavior on social media platforms. Essay 3 seeks to examine the individuals' information verification behavior on social media through the lens of individuals' fake news awareness, perceived cost of information verification, trust in social media and truth-seeking.

Kapitel 1: Im Rahmen des Erwartungsnutzenmodells leite ich ein theoretisches Modell von choice bracketing aus zwei verhaltensökonomischen Axiomen ab. Das erste etabliert einen direkten Zusammenhang zwischen narrow bracketing und correlation neglect. Das zweite identifiziert den Referenzpunkt als den Ort, an dem broad und narrow Präferenzen miteinander verbunden sind. In meinem Modell ist der narrow bracketer durch die Unfähigkeit, Veränderungen vom Referenzpunkt in unterschiedlichen Dimensionen gleichzeitig zu verarbeiten, charakterisiert. Kapitel 2: Warum geben Menschen, wenn man sie fragt, präferieren aber, nicht gefragt zu werden, und nehmen sogar, wenn sich die Gelegenheit ergibt? Wir zeigen, dass Axiome wie Separabilität, narrow bracketing, und scaling invariance diese scheinbar widersprüchlichen Beobachtungen vorhersagen. Insbesondere implizieren diese Axiome, dass die Interdependenz von Präferenzen (“Altruismus”) ein Ergebnis des Interesses für das Wohlbefinden anderer im Gegensatz zu ihren bloßen Auszahlungen ist. Hierbei wird das Wohlbefinden durch die referenzabhängige Wertfunktion aus der Prospekttheorie erfasst. Kapitel 3: Wir untersuchen, wie sich fake news auf den Informationsfluss zwischen Nachrichtenportalen und  ökonomischen Agenten auswirkt. Wir erweitern das klassische cheaptalk- Modell um Unsicherheit über die Präferenzen des sender (Nachrichtenportal). Es gibt zwei Typen von Nachrichtenportalen. Ein fake-news-Portal möchte im Agenten unabhängig vom wahren Zustand eine maximale Erwartung wecken. Ein legitimes Nachrichtenportal möchte die Wahrheit offenbaren. Wir zeigen, dass jedes informative perfekte Bayesianische Gleichgewicht durch einen Schwellenwert charakterisiert ist. Während der Agent alle Zustände unter dem Schwellenwert unterscheiden kann, ist es ihm unmöglich, Zustände über dem Schwellenwert zu unterscheiden. / Chapter 1: I derive a theoretical model of choice bracketing from two behavioral axioms in an expected utility framework. The first behavioral axiom establishes a direct link between narrow bracketing and correlation neglect. The second behavioral axiom identifies the reference point as the place where broad and narrow preferences are connected. In my model, the narrow bracketer is characterized by an inability to process changes from the reference point in different dimensions simultaneously. Chapter 2: Why do people give when asked, but prefer not to be asked, and even take when possible? We show that standard behavioral axioms including separability, narrow bracketing, and scaling invariance predict these seemingly inconsistent observations. Specifically, these axioms imply that interdependence of preferences (“altruism”) results from concerns for the welfare of others, as opposed to their mere payoffs, where individual welfares are captured by the reference-dependent value functions known from prospect theory. The resulting preferences are non-convex, which captures giving, sorting, and taking directly. Chapter 3: We present a theoretical model to investigate how the presence of fake news affects information transmission from media outlets to economic agents. In a standard cheap talk framework we introduce uncertainty about the sender’s (media outlet’s) preferences. There are two types of media outlets. A fake news outlet wants to push the agent’s belief to the maximum irrespective of the state of the world. A legitimate outlet wants to reveal the true state to the agent. We show that any informative perfect Bayesian equilibrium of our game is characterized by a threshold value. While the agent can perfectly separate amongst states below the threshold value, there is no separation amongst states above the threshold value. We determine the unique most informative threshold value for a general class of equilibria.

No description available.

In an attempt to create a fake news detection tool using a large language model (LLM), the emphasis is on validating the effectiveness of this approach and then making the tooling readily available. With the current model of gpt-4-turbo-preview and its assistant capabilities combined with simple prompts tailored to different objectives. While tools to detect fake news and simplify the process are not new, insight into how they work and why is not commonly available, most likely due to the monetization around the current services. By building an open-source platform that others can expand upon, giving insight into the prompts used, and enabling experimentation and a baseline to start at when developing further or taking inspiration from.  The results when articles are not willfully written as fake but missing key data are obviously very hard to detect. However, common tabloid-style news, which are often shared to create an emotional response, shows more promising detection results.

No description available.

As with every U.S. election cycle, the early 2016 election season was a roller-coaster experience, with positive and negative campaign messages emerging from both parties, and with high hopes and dashed hopes for both Democratic and Republican candidates. However, as Donald J. Trump emerged as the Republican candidate to challenge Democrat Hillary R. Clinton for the Office of President, the United States appeared to be equally bemused, horrified and confident in a Democratic victory — after all, how could someone so unorthodox as Donald Trump become the next president? His divisive campaign had become predictive, for many voters, of a clear victory for Clinton. However, those within vulnerable groups, including women, immigrants and the LGBTQ communities, along with many allies, noted the growing normalization of his micro- and macro-aggressions by the mainstream media, and the societal legitimization of his campaign.

A form of political disinformation that had a prominent place in journalism during the US presidential election in 2016 was called fake news. There were, however, uncertainties what could be called fake news. The purpose of this study was to explore qualitatively how mainstream media described the word fake news. The data collection approach was inductive. Data were collected primarily by searching the electronic media databases. The qualitative analysis of 212 articles resulted in two overarching themes which outlines how the concept of fake news has been described in meanstream media. The results of the analysis resulted in the two themes: fabricated lie and distorted truth. The main conclusion is that it circulates two different meanings for the concept of fake news. These definitions of fake news have different meanings, areas of action, actors behind och motives.

This study takes a qualitative approach to contextualizing and examining the communication of the so-called Kremlin trolls on Twitter, in relation to two major political processes that occurred in 2016, namely the Brexit referendum and the U.S. presidential election. Moreover, the study examines the news media of the two respective countries’ portrayal of the “Kremlin trolls”. The study assesses and problematizes mainstream application and contemporary usage of terminology in relation to two phenomena central to this thesis, namely: “Kremlin trolls”, and “fake news”. The study reconceptualizes the respective concepts into the “Kremlin meddlers” and “deceitful news”, as it was found those terms more accurately reflect both phenomena. Two sets of empirical data are examined in the thesis, one of which consists of 62 posts derived from 14 accounts of the Kremlin meddlers’ Twitter accounts. The other consistsof 30 articles stemming from 10 different news outlets, 5 of which were U.K. news outlets and 5 U.S. news sites, covering the political spectrum from one end to the other. The study applies one theoretical framework toexamine both sets of empiricaldata, namely Teun van Dijk’s Critical Discourse Analysis. The Kremlin meddlers’ Twitter accounts used two strategies to influence the electorates of both countries, one of which was to hide under the guise of pseudonyms impersonating actual citizens of the two nations. The other strategy employed by the meddlers was to maintain accounts that simulated news outlets, acting and appearing much like a legitimate news outlet would on the platform. The communication of the Kremlin meddlers was primarily aimed atsupporters of the ideological right by (re)producing discourse highly critical of the ideological left, this is particularly so for the Twitter accounts seeking to impersonate actual people. The pursuit of ideological polarization is centralin their communication.In the news media’s portrayal of the meddlers,differences are foundacross the ideological spectrum. The study identifies three themesin the discourse; one portraying the meddlers’ in a humanizing view, one creating a Them vs. Us categorization between Russia and the West, and the last focusing on the meddlers’ impacton democratic processes. Intrinsic to all of the themes is the (re)production of elite discourse, primarily seen through the selective use of voices from the symbolic elite to construct the social reality.