Procedural fairness, accountability, and legitimacy in the AI-enabled state: evidence from China
The study is persuasive evidence about how perceived legitimacy forms around AI-enabled state action, but it cannot establish legitimate authority because acceptance, fairness perception, and accountability perception are not substitutes for contestable institutional responsibility.
Review
Fan and Lyu examine whether AI-mediated state action is accepted through the same legitimacy mechanisms that structure conventional public authority. Their model separates perceived procedural fairness, perceived AI accountability, trust in state institutions, perceived legitimacy of the AI-enabled state, AI literacy, and acceptance of AI-embedded legal duties. A survey of 642 respondents and 41 semi-structured interviews across three major Chinese regions is analysed using PLS-SEM and qualitative interpretation. The reported relationships place legitimacy in the middle of the causal story: procedural fairness and perceived accountability increase perceived legitimacy, legitimacy strongly predicts acceptance, and institutional trust operates primarily through legitimacy rather than as an independent route to acceptance.
This matters because the paper resists a common governance shortcut in which technical performance, institutional trust, or system use is treated as sufficient evidence of legitimate authority. Its respondents also express conditional support, pairing efficiency and consistency benefits with concerns about opacity, data misuse, and reduced human oversight. The study therefore gives empirical support to the proposition that AI adoption by the state remains institutionally mediated rather than becoming legitimate by virtue of automation.
The boundary of that evidence is consequential. The study measures perceptions of accountability and legitimacy, not enforceable responsibility, legal authority, accessible appeal, correction, or redress. Acceptance of an AI-mediated duty can therefore indicate that a procedure is seen as appropriate without establishing that the underlying decision-maker is entitled to impose it or that affected people can contest error or abuse. The geographically concentrated, centralized governance context also limits claims beyond the studied population. The paper is most useful when read as evidence about legitimacy formation, not as proof that perceived legitimacy validates the institutional arrangement itself. My assessment is therefore narrower than the paper's broader legitimacy framing: it shows how an AI-enabled state may become accepted, not when its authority becomes justified. That judgment would change only with evidence connecting perceived accountability to enforceable responsibility, accessible contestation, and actual institutional constraint.
Key Insight
The study is persuasive evidence about how perceived legitimacy forms around AI-enabled state action, but it cannot establish legitimate authority because acceptance, fairness perception, and accountability perception are not substitutes for contestable institutional responsibility.