@article{DU2026, 
author = {Yabin DU},
title = {The Impact of Relative Performance Information on Public Trust in AI-enabled Public Services},
year = {2026},
journal = {China Public Administration Review},
volume = {8},
number = {2},
pages = {147-172},
keywords = {AI-enabled public services, relative performance information, message-sidedness, cognitive bias, trust in government},
url = {https://www.sciopen.com/article/10.26599/CPAR.2026.9680207},
doi = {10.26599/CPAR.2026.9680207},
abstract = {The proliferation of AI-enabled public services has reshaped administrative processes and citizen-state interactions, yet public trust in such services remains ambivalent. This study examines how message-sidedness—i.e., whether performance information about AI-enabled public services presents only advantages (one-sided messages) or both advantages and disadvantages (two-sided messages)—influences public trust in the services and government agencies. Drawing on the real-world case of the “instant approval” reform for individual business registration in China, we conducted two survey experiments using traditional manual service as the benchmark.Our empirical findings are threefold. First, message-sidedness significantly affects the persuasive power of relative performance information. One-sided messages, which only highlight efficiency gains of AI services, fail to significantly alter trust in either the services or the government. In contrast, two-sided messages—reporting both shorter processing times and decreased accuracy—significantly reduce trust, suggesting a potential negative bias effect.Second, we identify the dual role of rational cognition and cognitive biases in citizens' performance information processing. Although two-sided messages directly reduce trust, they also improve perceived performance knowledge, which indirectly enhances trust. This mediation effect indicates that rational cognition may partly offset the detrimental impact of negative disclosures. Thus, public trust in AI-enabled services is jointly shaped by reasoned judgment and affective heuristics.Third, we examine confirmation bias via citizens' pre-existing attitudes toward artificial intelligence. Those with higher optimism toward AI are more receptive to positive one-sided messages and less reactive to negative content in two-sided messages. This confirms that prior beliefs systematically moderate the impact of performance communication, underscoring the relevance of tailored messaging. Interestingly, when AI optimism is operationalized specifically as support for government adoption of AI, it paradoxically weakens the trust-enhancing effect of one-sided messages—likely due to heightened expectations and greater disappointment in the face of incomplete information.These results contribute to the growing literature on behavioral public administration by integrating performance management, trust theory, and information processing perspectives. We extend prior research by unpacking how the format and framing of relative performance information, rather than its valence alone, condition citizens' evaluations of AI-enabled services.Practically, the findings imply that governments should abandon one-sided, overly promotional communication strategies when introducing AI reforms. Instead, phased and audience-sensitive messaging—highlighting strengths while transparently acknowledging risks—can both foster trust and prevent future backlash. Moreover, enhancing digital literacy and addressing variation in AI attitudes across subgroups will be essential to securing public legitimacy in the era of algorithmic governance.}
}