Assist or substitute? The influential mechanism of worker's usage experience on their overall evaluation of generative artificial intelligence

Generative artificial intelligence (generative AI) plays a vital role in developing productivity, while also reshaping the way workers work and bringing about career shock. This research aims at enriching the understanding of factors influencing workers' attitudes toward generative AI and its underlying mechanism. According to elaboration likelihood model (ELM), usage experience as elaborated information is processed through the central route, shaping attitudes and perceptions. We conducted five studies including experiments and questionnaire surveys. The results demonstrate that: (1) worker's usage experience of generative AI is positively related to their overall evaluations of generative AI; (2) the relationship between usage experience and overall evaluation is mediated by assist-perception rather than substitute-perception; (3) creative self-efficacy can moderate the relationship between usage experience and overall evaluation, as well as the indirect effect path of assist-perception; (4) Employability can moderate the relationship between usage experience and overall evaluation. For workers with low creative self-efficacy and low employability, usage experience does not improve their overall evaluation. Theoretically, this study extends the understanding of antecedents that shape workers' attitudes toward generative AI and identifies the relative independence of perceived assistance and substitution. It practically offers managerial recommendations for addressing the opportunities and challenges posed by generative AI. Future research may build on this work by further exploring how usage experience influences perceptions across different technological and occupational contexts.