Answers are grounded in this paper and cite their sources.
Observed effects of large language models on job insecurity and turnover in German white-collar workers, indicating minimal impact.
Large language models like ChatGPT dominate the public discourse with their potential to take over human labor. ChatGPT already outperforms human employees in specific cognitive tasks in white-collar jobs, potentially substituting work tasks. However, whether employees perceive ChatGPT as threatening to their employment and how they intend to react to such potential threat remains largely unclear. Drawing from the conservation of resources theory and recent empirical findings on embodied robots, we investigate whether mere exposure to ChatGPT (Study 1) or use of ChatGPT (Study 2) evokes affective automation-related job insecurity and, subsequently, turnover and learning intentions. Further, we investigate the moderating role of core self-evaluations. The results of two online-experiments with German white-collar workers (N1 = 254; N2 = 391) demonstrate that neither mere exposure to ChatGPT nor use of ChatGPT elevate affective automation-related job insecurity, turnover and learning intentions. Findings from an exploratory analysis in Study 2 support that white-collar workers who perceive artificial intelligence to exceed their performance report higher affective automation-related job insecurity after the use of ChatGPT. Overall, the results do not provide empirical evidence of increased affective automation-related job insecurity and turnover and learning intentions, regardless of participants’ level of core self-evaluations. Thus, they do not support the transferability of positive effects from embodied robots on job insecurity to large language models.
No takes yet. Share an insight, caveat, or question.
Walczok et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: