Delegating Authority to Algorithms: Legitimacy and Public Values in European Tax Administrations

This paper develops a philosophical and empirical framework for evaluating the democratic legitimacy of algorithmic decision-making in European tax administrations. Drawing on evidence from fifteen jurisdictions, it investigates how four core public values—intelligibility, contestability, non-domination, and equality before the law—are articulated in policy discourse and translated into institutional practice within AI-enabled tax governance. Using a large corpus of legal texts, policy strategies, and oversight reports, the study codes qualitative insights into a structured binary feature matrix and analyses this dataset using Multiple Correspondence Analysis and hierarchical clustering. The results reveal four recurrent constellations of algorithmic governance: pragmatic enforcement, rights-anchored governance, constitutional caution, and data-intensive profiling, with Finland exhibiting a distinctive pattern of constitutional and procedural restraint. These empirical configurations form the foundation for the paper’s central philosophical question: Under what conditions can discretionary authority be legitimately delegated to algorithms? The analysis shows that although European administrations routinely appeal to ethical principles and human-centred AI rhetoric, many implementations transform these principles into managerial slogans, thereby weakening procedural justice, diminishing intelligibility, and limiting avenues for contestation. The paper argues that algorithmic decision-making can be democratically legitimate only when it sustains the four public value conditions that anchor administrative authority. It concludes by proposing a reconceptualization of algorithmic legitimacy as the preservation of intelligibility, contestability, non-domination, and equality within computationally mediated public administration.