Skip to main content

Open Access 2025 | Open Access | Buch

Automated Detection of Media Bias

From the Conceptualization of Media Bias to its Computational Classification

insite
SUCHEN

Über dieses Buch

Dieses Open-Access-Buch untersucht die automatisierte Erkennung von Medienvoreingenommenheit, wobei der Schwerpunkt auf Voreingenommenheit durch Wortwahl in digitalen Medien liegt. Die zunehmende Verbreitung digitaler Informationen bietet Chancen und Herausforderungen für die Analyse von Sprache, wobei kulturelle, geografische und kontextuelle Faktoren die Darstellung von Inhalten beeinflussen. Trotz des interdisziplinären Charakters der medialen Voreingenommenheitsforschung in Bereichen wie Linguistik, Psychologie und Informatik wird das Problem in bestehenden Arbeiten häufig aus begrenzten Perspektiven angegangen, da es an umfassenden Rahmenwerken und zuverlässigen Datensätzen mangelt. Das Buch zielt darauf ab, diese Lücken zu schließen und einen systematischen Ansatz zur Erkennung von Medienvoreingenommenheit vorzuschlagen. Es entwickelt funktionsbasierte und vertiefende Ansätze zur automatisierten Bias-Erkennung, darunter ein BERT-basiertes Modell und MAGPIE, ein Multi-Task-Lernmodell. Diese Methoden zeigen eine verbesserte Leistung bei etablierten Benchmarks und zeigen das Potenzial tiefen Lernens bei der Erkennung von Medienvoreingenommenheit. Schließlich befasst sich der Autor mit den praktischen Anwendungsmöglichkeiten automatisierter Voreingenommenheit, wie der Verbesserung der Nachrichtenlektüre durch Vorwarnmeldungen, Textanmerkungen und politische Klassifikatoren, und untersucht die Auswirkungen von Voreingenommenheit auf das Engagement in den sozialen Medien.

Inhaltsverzeichnis

Frontmatter

Open Access

Chapter 1. Introduction
Abstract
This thesis addresses the issue of automatically identifying media bias, mostly linguistic bias, in news articles. A varying word choice in any news content may have a major effect on the public and individual perception of societal issues, especially since regular news consumers are mostly unaware of the degree and scope of bias. Detecting and highlighting media bias is generally a challenging task since it is context-dependent, can be expressed in many ways, and its perception even differs based on personal perception and background.
Timo Spinde

Open Access

Chapter 2. Media Bias
Abstract
As portrayed in Chap. 1, media bias is a complex concept to identify and analyze. To construct a coherent framework to cover different bias types and to understand the state of the art in computer science when dealing with the domain of media bias is crucial to contribute meaningful research. To do so, in this chapter, we give a more detailed overview about media bias theory, as well as provide the results of an extensive literature review on automated media bias detection methods. To start the journey into this thesis here from a common ground zero, we will briefly summarize the general background of bias first.
Timo Spinde

Open Access

Chapter 3. Questionnaire Development
Abstract
As we laid out in Chap. 2, media bias is a complex concept to identify and analyze. Constructing a coherent framework to cover different bias types has been an important step in setting a basis for future work. However, not only inconsistent definitions of media bias have called for a unified approach; previous assessment strategies of bias are similarly lacking overlap and empirical evaluation.
Timo Spinde

Open Access

Chapter 4. Dataset Creation
Abstract
At this juncture, we have provided an overview of the media bias domain in Chap. 2, and we have explored how to query the perception of bias in Chap. 3. Nevertheless, we also witnessed the vast array of methods for measuring media bias. This naturally results in an extensive assortment of datasets within the domain.
Timo Spinde

Open Access

Chapter 5. Feature-based Media Bias Detection
Abstract
Thus far, we have presented a comprehensive literature review on media bias in Chap. 2, evaluated reliable measures for understanding media bias perception in Chap. 3, and introduced our two new datasets, MBIC and BABE, in Chap. 4. We now turn our attention to the design and implementation of automated bias classification systems. This chapter centers on a traditional machine-learning approach grounded on linguistic features.
Timo Spinde

Open Access

Chapter 6. Neural Media Bias Detection
Abstract
We already introduced existing methodologies to automatically detect media bias in Chapter 2. We also saw one major example of a traditional feature-based classification system in the previous Chapter 5. Now, we dive more into neural classification within the domain, particularly the major approaches developed during this dissertation.
Timo Spinde

Open Access

Chapter 7. Visualization and Perception of Media Bias
Abstract
As we discuss throughout the thesis, media bias plays a significant role in shaping individual and collective perceptions of news, which can have profound implications on how people form opinions and make decisions based on the information they consume. In order to address the potential negative consequences of media bias, it is crucial to explore effective communication strategies that can counteract its effects. However, existing research on the theoretical foundations of bias messages and visualizations is limited, with neither visualization theory nor bias theory providing comprehensive insights into effective approaches for addressing this issue within their respective domains.
Timo Spinde

Open Access

Chapter 8. Conclusion and Future Work
Abstract
This chapter summarizes and concludes the contribution of this thesis in Sect. 8.1 and Section Sect. 8.2, respectively. Sect. 8.3 provides an overview of future work projects and the ethical implications of working on media bias.
Timo Spinde
Backmatter
Metadaten
Titel
Automated Detection of Media Bias
verfasst von
Timo Spinde
Copyright-Jahr
2025
Electronic ISBN
978-3-658-47798-1
Print ISBN
978-3-658-47797-4
DOI
https://doi.org/10.1007/978-3-658-47798-1