The Past as a Distant Mirror

University and college professors have been using simulations for teaching and research purposes for some time now, at least since the early 1980’s when personal computers made their first appearance. We’ve come a long way. The crude images of that era have been replaced with photo-realistic simulations.

Playing games, however, differs significantly from simulations meant to educate and inform. With cultural and historical simulations, special attention must be given to how the environment is constructed. Accuracy is important. As my historian friend Will Hasty puts it, the underlying issue is that of alterity. This is the word professional historians use to highlight the “otherness” of historical events and periods. Consider the 4th Crusade, which ended with the sacking of Constantinople in 1204. A historian can attempt to reconstruct that world and the thoughts and ideas of those who lived in it, but the final result is always partial and incomplete. That’s because the world of the 13th century is far removed from our own. The noted historian Barbara Tuchmann captured the essence of alterity in the title of one of her books: A Distant Mirror.

All historical events are distant to one degree or another. Even a reconstruction of what I did yesterday or the day before is partial and incomplete. Add 800 years to the mix, and a rough approximation is the best a historian can achieve. An even greater challenge is the all too human tendency to read our present back into the past. The historians even have a word for this too. It’s called presentism, the interpretation of past events in terms of present values and concepts. One can never entirely eliminate present understandings or assumptions while constructing historical narratives, and those who think they have are delusional. Or as my colleague Will Hasty puts it, “Scholars such as Gadamer and Ricoeur have argued in different ways that historical understanding is always mediated by present-day assumptions: representations of the past are not transparent windows but constructed interpretations.” ¹ This doesn’t mean all historical truth is relative. Historical research methods exist for a reason. When properly employed, they act as guardrails, ensuring that one keeps their “presentness” in check.

The addition of generative AI complicates this story even more. The underlying problem is that current models are trained mainly on modern data. Just as a human absorbs a present-day orientation through reading and cultural interactions mediated by words, our AI models acquire a similar orientation by ingesting billions of modern documents. The seemingly effortless response to our historical prompts can lead to the mistaken idea that the model is right when, in fact, it may be doing what humans often do – generating a narrative driven by modern concerns masquerading as the fruit of a genuine historical inquiry.

The larger concern for historians is that we will begin to accept these AI-generated responses and simulations without ever questioning them. We will abandon the work of critical engagement, preferring instead the gratification of the immediate. This concern is even more pressing when the digital worlds we create become photo-realistic. If seeing is believing, will vision alone be the final arbiter of what’s true in this emerging world? Questions like these highlight the importance of striving to achieve historical fidelity in everything we do with AI. Our AI-enabled historical simulations and digital worlds ought to be accurate, but they should also remain “other” to the world we now live in. Or, as my friend Will writes, let’s “not collapse or ‘flatten’ our distance from the past over-hastily.” ² Alterity might be a 25-cent word, but it has earned its keep in the age of AI.

1. Hasty, Will. “True Digital Poiesis: Generative AI, Historical Simulation, and the Renegotiation of Alterity.” (Unpublished manuscript, May 2025): 2.

2. Hasty, “True Digital Poiesis…” (Unpublished manuscript, May 2025): 3.