A critical look at the epistemological and technological barriers to building AI-based historical personas.
The idea of creating an AI chatbot that fully replicates a historical figure is both fascinating and problematic. Despite significant advancements in artificial intelligence and natural language processing, the creation of a fully authentic historical persona remains epistemologically and technologically impossible. Several reasons support this conclusion:
1. The data just isn’t all there
A lot of potentially valuable documents have been destroyed, lost, or simply never written. What we do have often shows just a few selected scenes from a much bigger picture – not enough to reconstruct a full, coherent personality.
2. No one’s life is fully recorded
Not even today. Most of our thoughts, feelings, and daily decisions never get documented – let alone those of someone who lived a century ago. If it never made it into the historical record, AI has nothing to work with.
3. Sources can’t always be trusted
Letters can lie. Memoirs can be censored. Even first-person accounts may reflect fear, vanity, or outside pressure. Just because something was written doesn’t mean it was true – or honest.
4. We can’t fully recover historical context
Past eras had their own codes, assumptions, and ways of thinking. Language carried different meanings. Behaviors we see as odd today may have been normal then. AI trained on modern data often misses these nuances – and that distorts the simulation.
5. People change – and AI doesn’t track that
Historical figures weren’t static. Their views evolved, their tone shifted, their temperament matured or broke down. Without detailed, long-term records, it’s impossible to recreate that psychological arc. AI usually offers a frozen snapshot, not a life in motion.
6. The past is filtered – sometimes heavily
What gets preserved in archives is often shaped by politics, ideology, or convenience. Key details may be erased or rewritten. Unless these filters are made explicit and corrected, they’re baked into the AI.
7. We see them through modern eyes
It’s hard not to. We tend to idealize or condemn historical figures based on our current values. That projection seeps into how we train and interpret AI models. The result isn’t a faithful mirror of the past – it’s a reflection of our present.
While a fully authentic AI recreation of a historical figure remains beyond our reach, this does not render the endeavor futile. What we can build – when grounded in verified sources, historically accurate language, and transparent boundaries – is a credible and meaningful simulation.
Such AI personas can serve as powerful educational tools, narrative devices, and platforms for public engagement with history. They allow us to explore how someone might have thought or spoken, within the limits of what is knowable.
Authenticity, in its absolute sense, may be unattainable – but historical fidelity and interpretive honesty offer a path forward. When developed responsibly, these AI models can deepen our understanding of the past, not by resurrecting it, but by illuminating it.