Social inequality in algorithmic societies : empirical studies on AI-driven technologies and the public

Szafran, Daria

Preview

PDF
Daria Szafran_Dissertation 2026_MADOC.pdf - Published
Download (5MB)

URN:	urn:nbn:de:bsz:180-madoc-724176
Document Type:	Doctoral dissertation
Year of publication:	2026
Place of publication:	Mannheim
University:	Universität Mannheim
Evaluator:	Kreuter, Frauke
Date of oral examination:	2026
Publication language:	English
Institution:	Außerfakultäre Einrichtungen > Mannheim Centre for European Social Research - Research Department A School of Social Sciences > Social Data Science and Methodology (Keusch 2022-)
Subject:	300 Social sciences, sociology, anthropology
Keywords (English):	artificial intelligence , algorithm , social inequality , algorithmic decision-making , smart city , social media , Instagram , body image
Abstract:	Technologies driven by artificial intelligence (AI) are increasingly embedded in many domains of social life, spanning institutional decision-making, urban governance, and everyday digital platforms. While AI technologies promise gains in efficiency and productivity, growing evidence suggests that they might come with the risk of reproducing and exacerbating existing social inequalities by disproportionately disadvantaging historically marginalized social groups. These developments highlight that AI technologies are not neutral tools but sociotechnical artefacts whose design, deployment, and consequences are shaped by broader social structures and power relations. Despite being directly affected by algorithmic outcomes, members of the public are often excluded from decisions about the development and governance of AI technologies. Drawing on constructivist and critical theoretical approaches, this dissertation investigates how members of the public perceive, evaluate, and are affected by AI technologies, and how these processes are structured by social inequality. The three empirical chapters focus on fairness perceptions of automated decision-making (ADM), attitudes toward smart city technologies, and the effects of social media platforms. The analyses draw on population-based probability samples from Germany and combine qualitative content analysis with quantitative methods, including latent class analysis and two-stage least squares estimation. The findings demonstrate that public understandings of fairness in algorithmic decision-making extend beyond technical definitions, placing particular emphasis on the role of human judgment and oversight. Furthermore, attitudes toward AI-driven smart city applications are systematically structured by socio-demographic characteristics and reflect broader patterns of social inequality. Finally, the results suggest that the consequences of digital platform use are heterogeneous across population groups, underscoring the importance of considering social context when assessing the impacts of technology. Taken together, the dissertation contributes to the sociological literature in the context of AI by demonstrating that the development, evaluation, and consequences of algorithmic technologies are inseparable from existing social inequalities. It argues that more inclusive approaches to technology governance are needed to ensure that those most affected by algorithmic systems are meaningfully represented in their design, deployment, and regulation.