Designing AI for humans: preventing manipulation and protecting digital agency

Paper by Melisa Basol: “Artificial intelligence (AI) is reshaping digital autonomy. This article examines AI-driven manipulation, exploring its mechanisms, ethical challenges, and potential safeguards. AI systems are increasingly integrated into personal, social, and political domains, shaping decision-making processes. While these systems are often framed as neutral tools, AI systems can manipulate on three levels: (1) structural manipulation, where AI systems shape decisions through design choices like ranking algorithms and engagement-driven models; (2) exploitation by external actors, where AI is leveraged to amplify the spread of harmful falsehoods, automated deception, and personalized manipulation attempts; and (3) emergent manipulation, where AI systems may exhibit unexpected or autonomous influence over users, even in the absence of human intent. The article underscores the lack of a cohesive framework for defining and regulating AI manipulation, complicating efforts to mitigate its risks. To counteract these risks, this article draws on psychological theories of manipulation, persuasion, and social influence while examining the power asymmetries, systemic biases, and epistemic instability that complicate efforts to safeguard digital agency in the age of generative AI. Additionally, this article proposes a tiered intervention approach at three levels: (1) capability-level safeguards; (2) human-interaction interventions; and (3) systemic governance frameworks. Ultimately, this article challenges the prevailing narrative of an AI ‘revolution’, arguing that AI does not inherently democratize knowledge or expand autonomy but instead risks consolidating control under the guise of progress. Effective governance must move beyond transparency towards proactive regulatory mechanisms that prevent manipulation, curb power asymmetries, and protect human agency…(More)”.