Performing Responsibility: Leadership Discourse and Accountability for Algorithmic Harm

Sammanfattning

Social media platforms have increasingly become embedded in our daily lives. They have undergone a transformation from participatory communities into sophisticated algorithmic systems optimized for engagement. Despite the growing evidence of societal harms, including political polarization, radicalization, and addiction, platform leaders continue to publicly claim responsibility, while documented harms continue. This thesis examines how responsibility is discursively constructed and managed regarding algorithmic harm, and why systemic contradictions between that discourse and societal realities persist despite public scrutiny. The study employs a qualitative case study of Mark Zuckerberg and Meta. Fairclough's critical discourse analysis model is applied to a corpus of Zuckerberg's congressional testimony, earnings call transcripts, and general public communications from 2018 to 2026. The analysis is guided by three theoretical frameworks, namely responsible leadership theory (Maak & Pless, 2006), surveillance capitalism (Zuboff, 2019), and legitimacy theory (Suchman, 1995). Three patterns emerge. Responsibility discourse is genre-governed and audience-specific, agency is systematically displaced when specific harms are examined, and the advertising revenue model is structurally excluded across every genre. These contradictions persist because they are structurally necessary for surveillance capitalism, discursively sustained through audience-calibrated legitimacy strategies, and normatively protected from the frameworks that critique them. The study concludes that closing the gap between performed and true accountability should not rest on individual leadership alone, but requires democratic intervention to impose it.