Weizenbaum Library (Weizenbaum Institute)
Not a member yet
864 research outputs found
Sort by
Power in AI and public policy
Full text linkThis contribution scrutinizes how AI is related to power, how power theory can contribute to the debate about public sector AI, and whether new concepts of power are needed to study AI. Based on the assumption that the current public debate about AI is a power struggle about who gets to set the rules for future societal development, this chapter draws from the rich legacy of theories of power and domination and develops a systematization of different conceptions of power that encompasses various ontological and dimensional distinctions of power, carving out their analytical foci, and related power struggles. The remainder of the argument scrutinizes how these conceptions of power relate to dominant discourses about the power of AI in public policy: the use of AI in public policy and its power implications, recent initiatives to regulate AI, and AI-triggered systemic criticism and propositions for new social orders and utopias
Algorithmically Curated Lies: How Search Engines Handle Misinformation about US Biolabs in Ukraine
Full text linkThe growing volume of online content prompts the need for adopting algorithmic systems of information curation. These systems range from web search engines to recommender systems and are integral for helping users stay informed about important societal developments. However, unlike journalistic editing the algorithmic information curation systems (AICSs) are known to be subject to different forms of malperformance which make them vulnerable to possible manipulation. The risk of manipulation is particularly prominent in the case when AICSs have to deal with information about false claims that underpin propaganda campaigns of authoritarian regimes. Using as a case study of the Russian disinformation campaign concerning the US biolabs in Ukraine, we investigate how one of the most commonly used forms of AICSs - i.e. web search engines - curate misinformation-related content. For this aim, we conduct virtual agent-based algorithm audits of Google, Bing, and Yandex search outputs in June 2022. Our findings highlight the troubling performance of search engines. Even though some search engines, like Google, were less likely to return misinformation results, across all languages and locations, the three search engines still mentioned or promoted a considerable share of false content (33% on Google; 44% on Bing, and 70% on Yandex). We also find significant disparities in misinformation exposure based on the language of search, with all search engines presenting a higher number of false stories in Russian. Location matters as well with users from Germany being more likely to be exposed to search results promoting false information. These observations stress the possibility of AICSs being vulnerable to manipulation, in particular in the case of the unfolding propaganda campaigns, and underline the importance of monitoring performance of these systems to prevent it
Individual heat adaptation: Analyzing risk communication, warnings, heat risk perception, and protective behavior in three German cities
Full text linkExtreme heat poses severe health threats, as the increased numbers of hospitalizations and fatalities during heat waves show, though little is known about adaptive behavior toward heat. We conducted a household survey on individual perceptions of heat stress and individual heat protection in the summer and autumn of 2019. In total, 1417 people from three medium‐sized German cities participated via telephone or online. Based on the Protective Action Decision Model (PADM), which we adapted to heat stress, we analyzed links between risk perception, environmental and demographic factors, perceptions of stakeholders, different heat warning messages, as well as actual and intended adaptive behavior. Overall, the PADM constructs explained around 16% of the variance in protection motivation, 19% in protective response, and 23% in emotion‐focused coping. Context factors (i.e., temperature, risk communication, gender, age, and homeownership) were significant predictors of the addressed outcome variables as were psychological factors (i.e., perceived personal vulnerability, response efficacy, response costs, preparedness, and perceived external responsibility). We further explored the effect of different warning messages on situational knowledge and intended behavioral adaptation in an experimental setting. Results showed that respondents felt significantly better informed after receiving a warning with action recommendations and reported more intended specific behaviors. Our research gives insights into individual protective action decision‐making processes. Based on our findings, we recommend tailoring risk communication strategies and combining heat warnings with action recommendations whenever possible to increase understanding and individual adaptation.Germany’s Federal Ministry of Education and Research (BMBF), FKZ 01LR1709A1 and 01LR2014
Potentials and Limitations of Active Learning: For the Reduction of Energy Consumption During Model Training
Full text linkThis article investigates the potential and limitations of using Active Learning (AL) to reduce AI’s carbon footprint and increase the accessibility of machine learning to low-resource projects. First, this paper reviews the recent literature on sustainable AI. The core of the article concerns AL as an emissions reduction technique. Because AL reduces the required data for model training, one can hypothesize that energy consumption and, accordingly, carbon emissions – also decreases. This paper tests this assumption. The leading questions concern whether AL is more efficient than traditional data sampling strategies and how we can optimize AL for sustainability. The experiments show that the benefit of AL strongly depends on its parameter settings and the data set size. Only in limited scenarios does the size reduction outweigh the computational costs for AL. For projects with more resources for annotations, AL is beneficial from an ecological perspective and should ideally be paired with model compression techniques. For smaller projects, however, AL can even have a negative impact on machine learning’s carbon footprint
Between control and participation: The politics of algorithmic management
Full text linkUnderstanding the role of human management is crucial for the debate over algorithmic management—to date limited to studies on the platform economy. This qualitative case study in logistics reconstructs the actor constellations (managers, engineers, data scientists and workers) and negotiation processes in different phases of algorithmic management. It offers two major contributions to the literature: (1) a process model distinguishing three phases: goal formation, data production and data analysis, which is used to analyse (2) the politics of algorithmic management in conventional workplaces, which differ significantly from platform companies. The article goes beyond surveillance to elucidate the role of the regulatory framework, various actors' knowledge contributions to the algorithmic management system, and the power relations resulting therefrom. While the managerial goals in the examined case were not oriented towards a surveillance regime, the outcome was nevertheless a centralisation of knowledge and disempowerment of workers
How COVID-19 and the News Shaped Populism in Facebook Comments in Seven European Countries: A Computational Analysis.
Full text linkCitizen-generated populism is flourishing in the comments sections of online news. The factors that shape the extent of such populist communication from below are still under-researched. This study focuses on the COVID-19 crisis to examine how contextual and media-related factors are related to the extent of populism in comment sections on Facebook pages of news outlets from seven European countries (AT, DE, FR, IT, NL, SE and UK). Computational text analysis, machine translation and Bayesian multilevel regression were used to analyze digital trace data from 65,258 posts and 3.4 million comments published between February 2020 and June 2021. The computational measurements - multilingual dictionaries for posts and distributed dictionary representation to capture populism in comments - were rigorously validated. The results show that posts referring to the government, experts, COVID-19, and restrictions exhibit higher levels of populism in the comments sections. The stringency of containment policies was positively associated with populism in Germany, Austria, and the Netherlands when COVID-19 was mentioned. Lower levels of populism were observed for tabloid media and when news outlets engaged in visible moderation. The implications of these findings beyond the pandemic context and methodological challenges are discussed.This research was supported by the Kaiserschild Foundation, the Austrian Federal Ministry of Education, Science and Research through an OeAD Marietta Blau-Scholarship [MPC-2021-02005], and the German Federal Ministry of Education and Research under grant number 16DII135
Notable enough? The questioning of women’s biographies on Wikipedia
Full text linkThis study focuses on biographies nominated for deletion in the German-language Wikipedia and the encyclopedia’s core principle of notability. Results are presented from quantitative content analyses of deletion nominations, discussions, and decisions from the year 2020. It shows that women’s biographies are more often called into question but not deleted more often than men’s biographies. Additionally, women’s biographies are discussed more controversially. Neither a lack of notability criteria, a lack of external sources, nor individual misogynistic users seem to cause this increased questioning. Instead, the results suggest that the notability of women is collectively surveilled and contested with higher intensity due to biased perceptions. This can be explained by the fact that the concept of notability is not value-free or gender-neutral in the first place—even though it is based on rational discourse. The gender gap in biographies is contentiously discussed by users themselves, too, while overt sexism and gender-based devaluations are effectively countered by engaged users.This work was supported by the German Federal Ministry of Education and Research (BMBF) under Grant 16DII125
Comparative Analysis of the Essential Factors for the Adoption of Massive Open Online Courses in Higher Education of a Developing Country: Pre and Post COVID-19
Full text linkAlthough massive open online courses (MOOCs) offer numerous benefits to students, developing countries are still in the early stages of promoting their implementation. This study aims to investigate how the factors influencing MOOC adoption have evolved in response to the increased usage of online courses during the pandemic. The proposed model is based on the Technology Acceptance Model, and research hypotheses are presented based on six different factors: Perceived Usefulness, Perceived Ease of Use, Openness, Self-Efficacy, Quality of Service, and Reputation of the MOOC Provider. To test these hypotheses two surveys were conducted, one before and one after the COVID-19 period. Analyzing the data from these two time periods provides insight into the level of influence each of these factors has had on increased MOOC usage. Survey data was tested using the novel Partial Least Squares-Artificial Neural Network approach, which can effectively analyze complex human decisions. The findings indicate that Perceived Usefulness was the most influential factor in the adoption of MOOCs both before and after the COVID-19 pandemic. Interestingly, changes have been observed in the impact of Openness between the pre-pandemic and post-pandemic periods.The Weizenbaum Institute is funded by the German Federal Ministry of Education and Research (BMBF
What is Next for Civic Design?
Full text linkThis conversation started from the observation that contrasting logics
between civic initiatives and institutional approaches often make it challenging for the
former to become sustainable and increase their impact. We therefore explored how
civic design could enhance the significance of local civic initiatives within institutional
settings. The vivid conversation rendered three key orientations in civic design: 1)
operating from within the community, 2) focusing on the interaction between civic
initiatives and governmental and academic institutions, and 3) taking a
transformational perspective on the interplay between civil society and its institutional
context in the information age. We identified prompting questions on four important
topics, primarily related to the second orientation: listening to citizens, fostering
collaborative relationships, ensuring continuity through funding, and scaling or
spreading local civic initiatives. These questions contribute to the agenda for next steps
on the role of civic design
Expanding Horizons or Eroding Human Competence?
Full text linkThe proliferation of generative AI (GenAI) applications in the workplace has led to widespread speculation about the future of work. In this discussion paper, we formulate five theses on the relationship between GenAI and work, based on theoretical considerations and initial empirical impressions. They also serve as hypotheses for the GENKIA research project, in which we empirically examine changes in work across programming, journalism, marketing, HR management and public administration.
The hypotheses are as follows:
(1) Despite technical breakthroughs, GenAI is not an equivalent to human intelligence;
(2) GenAI becomes usable through human labor;
(3) GenAI represents a new quality of interaction between humans and machines;
(4) The introduction of GenAI creates work;
(5) Generative AI requires new answers to ensure good working conditions