Unpacking the heterogeneity of pre-service teachers’ ChatGPT acceptance: a latent profile analysis across STEM and non-STEM disciplines

Pre-service teachers’ acceptance of generative AI is highly heterogeneous, yet most studies rely on variable-centered approaches that cannot reveal qualitatively distinct subgroups. This study adopted a person-centered latent profile analysis (LPA) to identify pre-service teachers’ ChatGPT acceptance profiles and to compare STEM and non-STEM teachers in terms of profile distribution and behavioral intention. Data were collected from N = 128 pre-service teachers in Taiwan (68 STEM, 60 non-STEM). Building on the Technology Acceptance Model (TAM) and Unified Theory of Acceptance and Use of Technology 2 (UTAUT2), five constructs were used as LPA indicators, with behavioral intention treated as a distal outcome. The analysis identified four highly distinguishable profiles (entropy = 0.985): Pragmatic Evaluators (47.66%), Environmental Observers (11.72%), Technology Pioneers (26.56%), and Resistant Skeptics (14.06%). Notably, Resistant Skeptics demonstrated relatively high perceived ease of use but very low behavioral intention, indicating that perceived technical ease does not guarantee behavioral intention. Disciplinary background was strongly associated with profile membership ( χ ² = 36.20, p < .001, Cramer’s V = 0.532), with STEM teachers concentrated in Technology Pioneers (47.1%) and non-STEM teachers overrepresented in Environmental Observers and Resistant Skeptics. The four profiles also differed substantially in behavioral intention ( F = 109.00, p < .001, η ² = 0.726). These findings demonstrate that person-centered approaches can uncover heterogeneous acceptance patterns and that disciplinary background systematically shapes profile membership. Practically, the study proposes a differentiated AI literacy training framework, emphasizing that profiles such as Resistant Skeptics require targeted interventions that extend beyond operational skills training.