Weizenbaum Journal of the Digital Society
Permanent URI for this collectionhttps://www.weizenbaum-library.de/handle/id/1115
The Weizenbaum Journal of the Digital Society (WJDS) is an interdisciplinary, diamond open access journal that investigates processes of digitalization in society from the perspectives of different research areas.
Browse
6 results
Search Results
Item Too Far Away from the Job Market – Says Who? Linguistically Analyzing Rationales for AI-based Decisions Concerning Employment Support(Weizenbaum Institute, 2024-07-03) Berman, AlexanderThis paper describes an AI-based decision-support system deployed by the Swedish Public Employment Service to assist decisions concerning jobseekers’ enrolment in an employment support initiative. Informed by previous research concerning explanations in relation to trust, appealability, and procedural fairness, as well as jobseekers’ needs and interests in relation to algorithmic decision-making, the study linguistically analyses the extent to which the system enables affected jobseekers to understand the basis of decisions and to appeal or take other actions in response to automated assessments. The study also analyses the degree to which rationales behind decisions accurately reflect the actual decision-making process. Several weaknesses in these regards are highlighted, largely resulting from the opacity of the statistical model and the linguistic choices behind the design of explanations. Potential strategies for increasing the explainability of the system as a means to meet the needs and interests of affected jobseekers are also discussed. More broadly, the study contributes to a better understanding of how the linguistic design of AI explanations can affect normative dimensions, such as trust and appealability.Item Can Sustainable Shopping Recommendations in Online Retail Help Reduce Global Warming?(Weizenbaum Institute, 2024-04-25) Hoffmann, Marja Lena; Nanevski, Ivana; Gossen, Maike; Bergener, Jens; Flick, Alexander; Santarius, Tilman; Biessmann, FelixTwo dominant and contradictory narratives describe the apparent contribution of information and communication technology (ICT) to climate change. On the one hand, ICT can reduce global greenhouse gas (GHG) emissions by, for example, supporting energy efficiency or promoting sustainable consumption. On the other hand, the increased energy demands of emerging software components leveraging artificial intelligence or machine learning can be directly and indirectly responsible for GHG emissions. This makes it critical to assess whether ICT mitigates or exacerbates net climate impacts and the contributing factors. The impacts of software have received relatively little attention and require the development of new approaches to conduct such assessments. In particular, the net effect of complex real-world applications is frequently not measured. In this study, we provide a detailed step-by-step assessment to quantify the net global warming potential of an online shopping recommendation system that encourages users to make sustainable consumption decisions. We consider the energy consumed and associated GHG emissions in the development and use of the software and compare these to the potentially avoided GHG emissions associated with more sustainable recommended options. The results demonstrate that the software has the potential to indirectly avoid more emissions than it causes and that changes at different steps of the software can amplify this.Item Potentials and Limitations of Active Learning: For the Reduction of Energy Consumption During Model Training(Weizenbaum Institute, 2024-04-08) Nenno, SamiThis 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.Item Editorial: Practicing Sovereignty – Interventions for Open Digital Futures(Weizenbaum Institute, 2023) Irrgang, Daniel; Herlo, BiancaThis issue is dedicated to the Weizenbaum Conference 2022, titled ‘Practicing Sovereignty: Interventions for Open Digital Futures.’ The Weizenbaum Institute’s annual gathering brought together researchers, networks, and collaborators to focus on the theme of ‘digital sovereignty.’ This term, hotly debated and used with varying connotations in fields such as research, activism, law, and policy-making, refers to competencies, duties, and rights in digital societies. The contributions compiled in this issue are based on papers presented at the 2022 conference. They explore notions of digital sovereignty in tension with topics such as AI deepfakes, algorithmic governmentality, ethics and datafication in the context of machine learning, and community-driven open-access publishing in academia.Item The Limits of Computation: Joseph Weizenbaum and the ELIZA Chatbot(Weizenbaum Institute, 2023) Berry, David M.Developed in the 1960s by Joseph Weizenbaum, ELIZA is arguably among the most influential computer programs ever written. ELIZA – and especially its most famous persona DOCTOR – continues to inspire programmers, wider discussions about AI, and imitations. This original ancestor of all conversa-tional interfaces and chatbots maintains a special fascination for engineers, historians, and philosophers of artificial intelligence (AI) and computing. With its ability to produce human-like responses using a relatively small amount of computer code, ELIZA has paved the way for a multitude of similar programs. These take the form of conversation agents and other human-computer inter-faces that have inspired entire new fields of study within computer science. This paper examines Weizenbaum’s contribution to AI and considers his more critical writings in the context of contemporary developments in generative AI, such as ChatGPT. Examining how ELIZA has been discussed can provide insights into current debates surrounding machine learning and AI.Item Making Arguments with Data: Resisting Appropriation and Assumption of Access/Reason in Machine Learning Training Processes(Weizenbaum Institute, 2023) Savic, Selena; Martins, Yann PatrickThis article presents an approach to practicing ethics when working with large datasets and designing data representations. Inspired by feminist critique of technoscience and recent problematizations of digital literacy, we argue that machine learning models can be navigated in a multi-narrative manner when access to training data is well articulated and understood. We programmed and used web-based interfaces to sort, organize, and explore a community-run digital archive of radio signals. An additional perspective on the question of working with datasets is offered from the experience of teaching image synthesis with freely accessible online tools. We hold that the main challenge to social transformations related to digital technologies comes from lingering forms of colonialism and extractive relationships that easily move in and out of the digital domain. To counter both the unfounded narratives of techno-optimismand the universalizing critique of technology, we discuss an approachto data and networks that enables a situated critique of datafication and correlationism from within.