AI in the Public Eye: A Randomized Survey Experiment of Citizens’ Legitimacy Perceptions of Automated Decision-making

This paper investigates the public's perceptions of legitimacy regarding automated decision-making in child welfare services. Understanding legitimacy perceptions is crucial as it mirrors the public's willingness to embrace AI in governance. We integrate institutional and psychological theories to explain the public's perceptions of legitimacy. Our research approach involves a randomized factorial survey experiment with a representative sample of the Finnish population, focusing on the impact of algorithmic disclosure and human involvement in decision-making in the context of automated needs assessments in child welfare services. This study builds on a previous identical survey experiment conducted with 932 Finnish civil servants, which found that both algorithm disclosure and human involvement considerably improved legitimacy perceptions. In the spring of 2024, we are extending this investigation to a citizen sample. While we anticipate that both algorithmic disclosure and human involvement will positively influence legitimacy perceptions in this group as well, we hypothesize that citizens may be more reluctant than civil servants to delegate decision-making to an algorithm, given the formers presumed lower levels of AI literacy. Conversely, we expect the public to place a higher value on algorithmic disclosure for legitimacy to strong normative connotations of transparency and openness, which civil servants might not be as susceptible to. Our research underscores the importance of algorithmic transparency and human interaction in the design and implementation of AI-driven governance arrangements.