Counterfactual History: Simulating Chinese Imperial Decisions with AI

This study employs a large language model (GPT-4o) to simulate the counterfactual decisions of three Chinese emperors: Emperor Huizong of the Song Dynasty, Emperor Chongzhen of the Ming Dynasty, and the late Qing imperial court in the nineteenth century. Drawing on bounded rationality theory (Simon, 1957) and prospect theory (Kahneman & Tversky, 1979), the study reassesses key decisions that led to dynastic collapse. The AI model proposes rational strategies such as negotiating with the Jurchens, appointing a crown prince early, or initiating institutional reforms. These strategies may have delayed or mitigated the decline. However, Confucian cultural norms and psychological pressures limit the feasibility of such solutions. The study highlights the potential of AI in historical analysis while also underscoring its limitations when confronting cultural and psychological constraints. It opens new directions for digital humanities research.