Abstract: This article explores how the widespread adoption of generative artificial intelligence, GAI, is reshaping the temporal fabric of postdoctoral research work, a segment of the scientific workforce that combines exceptionally high exposure to emerging automation with limited employment security and strong performance surveillance. Building on the theory of technological change, the study links substitution and complementarity channels to both objective and subjective measures of working time. Empirically, it draws on individual-level data from the 2023 Nature Global Post?Doctoral Survey, which collected responses from 3?715 postdoctoral researchers in ninety?three countries. To address the endogeneity of GAI uptake, the study employs two?stage least squares estimation, using cross?national variation in OpenAI regional access restrictions as an instrument, and triangulates the results with causal forest and Deep IV estimators in order to accommodate non?linear effects and complex heterogeneity. Across all methods, intensive GAI use is associated with statistically and substantively significant increases in weekly non?contract hours and overtime hours, averaging between 0.6 and 1.3 additional hours, along with discernible declines in reported time satisfaction and work–life balance. Mechanism analyses indicate that these outcomes arise primarily through a process of deskilling and task fragmentation rather than through productivity?induced task creation or wage gains. Specifically, the automation of routine scholarly functions, such as text drafting and coding, transfers effort toward engineering, iterative proofreading, and compliance checking, thereby reducing professional autonomy and increasing temporal fragmentation. Complementarity channels deliver few short?run benefits, as indicated by lower effective hourly earnings and muted growth in opportunities to initiate self?directed projects. Institutional context modulates the magnitude of these effects. In laboratories that exhibit entrenched overtime norms, the GAI-induced time burden is forty to seventy per cent larger, whereas in departments that report difficulty recruiting postdoctoral fellows, an indicator of stronger local bargaining power, the negative effects on both hours and time satisfaction are attenuated. Further heterogeneity analysis reveals an emergent distributional divide. Male researchers and individuals with above?median prior productivity incur smaller time penalties and, in some cases, modest efficiency gains, implying that current technological dividends accrue disproportionately to already advantaged actors and may widen inequality in early academic careers. The study contributes to debates on automation and academic labor by demonstrating that cognitive technologies can intensify rather than relieve time pressure even among high?skill workers, extending the insight of the industrious revolution literature to contemporary knowledge production. It also shows that organizational governance and labor market structure critically shape whether potential efficiency gains are converted into genuine leisure. Policy implications include the need to recalibrate performance metrics that encourage perpetual acceleration, to strengthen postdoctoral bargaining positions through transparent career pathways and collective representation, and to ensure inclusive access to high?quality AI resources and training.

摘要： 生成式人工智能（GAI）技术的迅猛发展，开启了从体力自动化向脑力自动化转型的劳动变革时代，科研生产模式与劳动时间结构正经历前所未有的冲击。研究聚焦博士后“一线”科研人员，探讨GAI对科研劳动时间的影响效应及其作用机制。研究基于系统性理论梳理与历史回顾，构建GAI形塑科研劳动时间的综合分析框架，利用Nature全球博士后调查数据，结合工具变量、因果森林与Deep IV等实证方法开展分析。结论显示，GAI深度应用显著延长了博士后的非契约劳动时间、超时劳动时间，并导致其时间满意度、时间平衡感同步下降。机制分析表明，GAI尚未引致生产率效应或新任务创造，反而通过“去技能化”进程加剧科研劳动时间负担。劳动过程控制、劳动议价能力对GAI的影响效应存在调节作用。未来应强化GAI应用监管，提升科研人员劳动自主性与岗位稳定性，推动GAI赋能科研创新。