Weizenbaum Library (Weizenbaum Institute)
Not a member yet
864 research outputs found
Sort by
A systematic review of echo chamber research: comparative analysis of conceptualizations, operationalizations, and varying outcomes
No full textThis systematic review synthesizes research on echo chambers and filter bubbles to explore the reasons behind dissent regarding their existence, antecedents, and effects. It provides a taxonomy of conceptualizations and operationalizations, analyzing how measurement approaches and contextual factors influence outcomes. The review of 129 studies identifies variations in measurement approaches, as well as regional, political, cultural, and platform-specific biases, as key factors contributing to the lack of consensus. Studies based on homophily and computational social science methods often support the echo chamber hypothesis, while research on content exposure and broader media environments, such as surveys, tends to challenge it. Group behavior, cultural influences, instant messaging platforms, and short video platforms remain underexplored. The strong geographic focus on the United States further highlights the need for studies in multi-party systems and regions beyond the Global North. Future research should prioritize cross-platform studies, continuous algorithmic audits, and investigations into the causal links between polarization, fragmentation, and echo chambers to advance the field. This review also provides recommendations for using the EU’s Digital Services Act to enhance research in this area and conduct studies outside the US in multi-party systems. By addressing these gaps, this review contributes to a more comprehensive understanding of echo chambers, their measurement, and their societal impacts
Ohne Intelligenz aber wirkmächtig: Künstliche Intelligenz als Akteur im öffentlichen Raum
Full text linkDas Thema Künstliche Intelligenz (KI) wird zurzeit unter anderem von der Frage bestimmt, wie die Ausbreitung algorithmischer Systeme die Öffentlichkeit verändert. Dahinter steht die Sorge vor Täuschung und Manipulation und vor Kontrollverlust gegenüber einer mächtigen Technik. Wenngleich weniger offensichtlich, ist jedoch auch die umgekehrte Blickrichtung interessant. Diese fragt danach, welchen Einfluss der öffentliche Diskurs auf unsere Wahrnehmung und unser Verständnis von KI hat. In den Kommunikationswissenschaften spricht man in diesem Zusammenhang von »talking AI into being«. Mit dieser schönen Formulierung verbindet sich die Beobachtung, dass öffentliche Diskussionen eine identitätsstiftende Quelle in der Entwicklung von Technologien sind, deren Einsatzmöglichkeiten und -grenzen noch im Dunkeln liegen
Dynamics of Data Work in AI Implementation Processes
Full text linkThe discourse on automation and artificial intelligence (AI) highlights the critical yet underexplored role of data work. While previous studies have focused on data work in the production of AI systems, this article investigates its significance in subsequent implementation phases. Drawing on two in-depth case studies from traditional industries in Germany – one in manufacturing and one in administration – we explore the labor and power dynamics inherent in these processes. Our findings reveal that AI implementation necessitates additional data work which cannot be outsourced due to its reliance on organization-specific knowledge. This dependency fosters new labor relations and power dynamics among development companies, organizational management, and workers, often leading to tensions and negotiation challenges. We introduce the concept of data work facilitation to describe the unique labor emerging from these dependencies and identify potentials for a new power resource workers could employ – data work bargaining power
Dokumentation zur Konferenz „Triggerpunkte: Anstöße für die politische Kommunikationsforschung“ mit Steffen Mau, Thomas Lux und Julian Heide. Weizenbaum-Institut, Berlin, 28. Juni 2024
Full text linkDie Dokumentation soll es ermöglichen, die Diskussionen auf der Konferenz „Triggerpunkte: Anstöße für die politische Kommunikationsforschung“ am 28. Juni 2024 im Weizenbaum-Institut nachzuvollziehen. Die Konferenz hatte das Ziel, politische Kommunikationsforscher:innen mit den Autoren der Studie „Triggerpunkte. Konsens und Konflikt in der Gegenwartsgesellschaft“ (Suhrkamp, 2023) ins Gespräch zu bringen. Diese Studie ist für die politische Kommunikationsforschung deshalb von Bedeutung, weil sie eine umfassende Kartierung der öffentlichen Meinung in Deutschland zu den wichtigsten politischen Themen mit bislang unerreichter Tiefenschärfe bietet. Dabei stellt sie einerseits klar, dass entgegen der medialen Darstellung in Deutschland ein hoher Konsens in politischen Grundsatzfragen herrscht. Andererseits macht sie deutlich, wie die öffentlichen Debatten, insbesondere in den sozialen Medien, die Polarisierung auch in Deutschland verschärfen – vorangetrieben durch „Polarisierungsunternehmer“.
Auf Seiten der Autoren nahmen Steffen Mau, Thomas Lux und Julian Heide teil. Erörtert wurden offene theoretische und empirische Probleme, etwa das Klassenkonzept oder die Erklärung für die Fehlwahrnehmung des Meinungsklimas. Vor allem aber wurden neun Ideen für mögliche kommunikationswissenschaftliche Forschungsprojekte diskutiert, die sich auf die „Triggerpunkte“ beziehen. Leitfragen der Ideen waren etwa: Welche Medienstimuli prägen die Wahrnehmung von Polarisierung und wie verändert dies die politische Kommunikation? Oder: Wie können stark affektiv geladene politische Debatten konstruktiv gewendet werden
Ethics of Data Work. Principles for Academic Data Work Requesters
Full text linkThe growing use of machine learning (ML) in academic research has led to a rising demand for large, labeled datasets. While the field initially relied on the labor of students and research assistants to label data, as models grew larger and more complex, there was a shift towards relying on large-scale, low-cost platforms like Amazon Mechanical Turk (MTurk) to label data at scale. However, this shift comes with serious ethical concerns. Now part of a massive industry, many data work companies exploit workers, leaving many workers facing low wages and precarious working conditions, with little institutional oversight or protection. Despite the centrality of this labor to modern research, ethical codes and guidelines from academic societies rarely address the implications of outsourcing data work to platform-based workers.
This paper advocates for the development of research ethics standards that ensure fair and responsible collaboration with data workers. We begin by defining the concept of “data work” and assessing how current ethical frameworks address it. We then highlight ongoing initiatives aimed at improving ethical regulation. Based on two focus groups and two expert workshops held at the Weizenbaum Institute in 2024, we propose a set of principles for academic data work requesters to guide ethical engagement with platform-based workers. Finally, we outline future steps for integrating these principles into scientific ethical codes and day-to-day research practices
Empowering smart regions: addressing challenges and leveraging enabling factors in municipal digital transformation
Full text linkThe advancing digital transformation of society creates a wide range of opportunities for improved access to information and resources that contribute to ensuring the availability of public services and the development of sustainable living spaces. This underlying potential does not only apply to urban areas; digitalisation projects are also being implemented in rural municipalities in order to exploit the potential of digital transformation. Nevertheless, the field of so-called smart regions has yet to receive substantial focus in research. To understand the specific challenges and enabling factors of digital transformation activities in urban-rural areas, a workshop was organised with the cooperation of municipal representatives of a model region in Schleswig-Holstein, Germany. Therein, specific technical, economic and social challenges as well as enabling factors of municipal digitalisation projects aimed at developing smart regions were identified. The results show that the success of digital transformation meas‐ ures in urban-rural areas is not only determined by the expansion of a corresponding technical infrastructure but in particular by the acceptance of citizens and municipality employees, as well as economic viability. This research further informs municipalities and future researchers on the critical factors required to effectively conduct digitalisation projects in the smart region context
Democratic Oversight of Government Hacking by Intelligence Agencies: A Critical Analysis of Brazil and Germany
Full text linkRegulating intelligence services is a challenge for modern societies worldwide. Their very modus operandi relies on tight secrecy protocols for the information gathered, internationally and domestically. Evolving surveillance techniques include exploiting the vulnerabilities of digital services, dealing on unregulated surveillance markets, and developing tailored tools. Theoretically, these actions aim at the public interest by, for instance, anticipating terrorist attacks or dismantling pedophilia networks. Nevertheless, they are increasingly used to surveil civil society without proper and demonstrated necessity or proportionality. Given the demand for increased transparency and accountability for intelligence agencies, especially when using hacking technologies, what institutional design and civic participation avenues for oversight may be proposed? How can (existing and yet-to-exist) institutions improve democratic external oversight activities in this realm? Through a comparison of Germany’s and Brazil’s legal frameworks and institutional ecosystems, the paper critically explores the meaning of “democratic oversight” of intelligence agencies, specifically observing oversight models for hacking operations. Looking at previous contributions by intelligence studies scholars in these countries and globally, the paper offers a critical-comparative analysis of institutional and political architectures to assess the levels of democratic participation. On this basis, it makes recommendations for both countries, which can be appropriated by external intelligence oversight bodies.The Weizenbaum Institute is funded by the German Federal Ministry of Education and Research (BMBF
Digital constitutionalism as an online speech governance framework: A critical approach
Full text linkThis chapter advances a critical approach to the theories of “digital constitutionalism”, in particular as a theoretical framework for recent initiatives targeting online speech governance. We build on previous work where we demonstrated overarching risks of borrowing from the symbolic load of the constitutionalist tradition to name and explain transnational normative phenomena that take place in private digitalised environments. We apply these critiques to the case of online speech governance by looking at two policy initiatives: the Meta Oversight Board, a private sector self-regulatory initiative implemented by the company Meta; and the European Digital Services Act. Our goal is to shed light on contradictions and misperceptions embedded in labelling online speech governance mechanisms as manifesta‐ tions of digital constitutionalism
How Do Environmental and Methodological Factors Influence Study Participants’ Answers in Surveys on Risk Perception in the Context of Climate Change and Heat Stress?
Full text linkResearch on climate change and impacts of natural hazards, such as heat waves, on human health has increased in recent years. Various approaches are used to study people’s attitudes and actions in this context, but little is known about the extent to which different modes or other environmental variables influence the results. Therefore, we ex- amined differences between surveys in three German cities, compared survey modes and investigated the influence of the temperature on the day of the survey and the previous days. We conducted two surveys on the topics of climate change risk perception and heat risk perception. In summer and autumn of 2019, in total 1,417 people from the three medium-sized German cities of Potsdam, Remscheid and Würzburg were surveyed via telephone or online. In sum- mer of 2020, 280 people were surveyed face-to-face in public parks in Potsdam. Climate change risk perception, the perception of heat waves as a health threat and the knowledge of heat warnings differed depending on place of resi- dence, survey mode and temperature. Participants of the online survey showed higher scores of risk perception than participants of the telephone and face-to-face surveys, indicating a self-selection bias. Increased temperature was associated with slightly higher levels of respondents’ heat wave risk perception and, among participants surveyed outside, climate change risk perception. The finding that both survey mode and environmental factors can influence survey results should be heeded when planning or interpreting and comparing studies
Lost in moderation: How commercial content moderation apis over- and under-moderate group-targeted hate speech and linguistic variations
Full text linkoai:weizenbaum-library.de:id/901Commercial content moderation APIs are marketed as scalable solutions to combat online hate speech. However, the reliance on these APIs risks both silencing legitimate speech, called over-moderation, and failing to protect online platforms from harmful speech, known as under-moderation. To assess such risks, this paper introduces a framework for auditing black-box NLP systems. Using the framework, we systematically evaluate five widely used commercial content moderation APIs. Analyzing five million queries based on four datasets, we find that APIs frequently rely on group identity terms, such as “black”, to predict hate speech. While OpenAI’s and Amazon’s services perform slightly better, all providers under-moderate implicit hate speech, which uses codified messages, especially against LGBTQIA+ individuals. Simultaneously, they over-moderate counter-speech, reclaimed slurs and content related to Black, LGBTQIA+, Jewish, and Muslim people. We recommend that API providers offer better guidance on API implementation and threshold setting and more transparency on their APIs’ limitations.Warning: This paper contains offensive and hateful terms and concepts. We have chosen to reproduce these terms for reasons of transparency