Abstract
When historians look back at past events, outcomes often appear inevitable. Equipped with hindsight, we naturally describe decisions in light of their eventual consequences. Yet for the people living through those events, the future was radically open, uncertain, and contested. How can historians recover that experience of contingency without surrendering empirical rigor? This talk introduces an experimental methodology that uses Generative AI and wargaming frameworks to model historical processes from the "inside out." Rather than employing artificial intelligence as an omniscient retrospective analyst, we use Large Language Model agents to simulate strictly bounded historical actors. By fencing each agent’s knowledge to a specific historical moment, role, and local information state, we strip away hindsight: the models deliberate forward into uncertainty without knowing how the story ends. Placed within structured, rule-governed wargame settings, these agents negotiate, miscalculate, and pursue conflicting intentions, generating emergent dynamics and unintended consequences. Using concrete examples, this talk demonstrates how combining natural-language AI with simulation environments offers historians a computational laboratory: a way to probe counterfactual possibilities, interrogate the causal weight of human agency versus structural constraints, and bridge the gap between the actor’s open horizon and the historian’s retrospective view.
Bio
William J. Turkel is a Professor of History at the University of Western Ontario and a former member of The College of New Scholars, Artists and Scientists of the Royal Society of Canada (2018-26). He received the Western Award for Innovations in Technology-Enhanced Teaching in 2021 and has been named one of the first cohort of Western's new Generative AI Teaching Fellows (2025-27). His latest monograph, on Exploring Global Twenty-First-Century History with Generative Artificial Intelligence, is currently under review at McGill-Queen's University Press.
