Social Sciences in China (Chinese Edition)
No. 8, 2026
The Historians That Enter the AI Laboratory: Practice and Methodological Reflections
(Abstract)
Gao Xi
Scholarly discussions of the relationship between AI and history have largely overlooked the structural role of engineers as a third party in the production of knowledge. Bringing this neglected role into the epistemological framework of AI history is one of the central issues in exploring “AI-historical thinking.” Experiments in which historians participated in the construction of a domain-specific AI Agent for History (HistAgent) and the benchmark evaluation framework (HistBench) show that historical knowledge must undergo sustained encounters and negotiations with the scientific logic of engineers, during which it is decomposed, recombined, and formalized through engineering translation before the model construction can be completed. This shows that technical discursive power over AI-historical thinking in practice remains largely with invisible engineers. The formation of AI-historical thinking depends on historians’ deep involvement in model design, corpus selection, and evaluation mechanisms. Although AI-historical thinking or logical reasoning engineered by technical teams follows pathways different from traditional modes of historical argumentation, this does not render it valueless. AI-historical thinking may, through sustained dialogue with historical traditions, inject an impetus that is difficult to foresee into the deeper transformations of historical paradigms.
