Hybrid images of generative AI: A Q methodological study of civil servants' perceptions

Abstract

The proliferation of Generative AI (GenAI) within public organizations has introduced agentic augmentation; however, its opaque and probabilistic outputs complicate accountability and routine bureaucratic processes. Despite the growing body of research on GenAI in government, the evidence on how civil servants interpret GenAI remains fragmented and is often examined through variable-centered acceptance models. To address this gap, we integrate four administrative logics within an Input-Process-Output framework, and employ Q methodology to systematically map the perceptions of 32 grassroots civil servants in China. The analysis yields four hybrid profiles—Cautious Institutionalists, Performance-Driven Optimists, Burden-Reduction Strategists, and Skeptical Followers—ranging from optimistic orientations emphasizing efficiency enhancement or workload alleviation, to more deliberative standpoints grounded in normative concerns, or pragmatic skepticism about technological readiness. These findings provide empirical evidence of the hybridity and complexity of civil servants' views, challenging and extending the sufficiency of utility-based technology acceptance models in the GenAI era. By highlighting how instrumental and normative rationalities intertwine, this study extends theoretical frameworks and offers holistic insights for navigating the complex adoption of GenAI in the public sector.