Artificial Intelligence & Historical Research

Writer: Marina Popovic, M.I

Introduction

In recent days it feels as if one cannot escape a day without hearing some kind of talk or news about artificial intelligence (AI)—from anything to do with Elon Musk, to generative AI being used to enhance Academy Award winner Adrien Brody’s Hungarian accent in The Brutalist, and cute AI images of animals popping up on social media feeds. Escaping AI feels almost impossible.

I began my educational career studying Medieval history, theology, music, and art. Afterwards, I attended library school and have since been working as a librarian for several years in the public sector. I have first-hand knowledge and experience about how important human labour is in our contextualization of the past and how we store information. Research is, and will continue to be, a critically human-specific job. So, once AI-generated images, videos, and even books began cropping up in my work, it made me immediately nervous for the future of art and intellectualism. The possibilities of AI are endless and it made me think: is AI the new “Big Bad” of the 2020s?

Not necessarily. 

Everything in this world has positive and negative aspects, and artificial intelligence is not an outlier to this rule. It all depends on the individual or organization who wields it. But how does AI pertain to today’s historical research? Does it help or hinder us in our understanding of the past? While I won’t be able to give a definitive answer to that question, and I doubt anybody can, I will analyze some current uses of AI in the historical research realm while examining both the pros and cons. It’s up to you, the reader, to form your own opinion of AI and its role in historical research.

So, What is Artitifical Intelligence?

AI is a form of “technology that enables computers and machines to simulate human learning, comprehension, problem solving, decision making, creativity and autonomy” [1]. It can run independently without the input of a human, can give recommendations, and can learn from new information. AI can be found in the seemingly simplest of technologies, such as Face ID on your phone to life-saving technologies found in hospitals. While AI has been a popular topic in recent media, it’s not a new technology. 

The history of artificial intelligence traces back to the 1950s when Alan Turing—famously known for decoding the German ENIGMA code during the Second World War—conducted a test by asking a person to distinguish between a human and a computer text response. The actual term “artificial intelligence” was coined in 1956 by computer scientist John McCarthy [2]. Throughout the remainder of the 20th-century, this emerging form of technology continued to grow, but it wasn’t until the early 2000s, once computer science entered a new era, that AI began to rapidly transform and evolve. 

Generative AI, the rising star of artificial intelligence technology, has been a driving force in modern popular AI usage. Programs like ChatGPT, DALL-E, Elon Musk’s Grok, and Microsoft Copilot can create text, images, sound, and videos based on prompts given by the user. Working as a librarian, I’ve experienced the rising usage of generative AI programs such as these in students submitting assignments, colleagues writing emails, creating programs for the public, and even something as simple as creating a silly image of a sloth reading a book to set as a computer background at work.

An AI generated image created using the prompt "Florence during the Italian Renaissance" on DeepAI, depicting white vans parked on the street during the Italian Renaissance.

The rising use of artificial intelligence in the last several years has raised questions about ethics and policy, as governance regarding usage is only just starting to emerge in institutions, companies, and governments. Research institutions, such as libraries, museums, universities, and disseminators of human knowledge are particularly important to think about when discussing AI usage. Recently, academic publisher Taylor & Francis sold access to research published through them to Microsoft for around $10 million without contacting authors about the AI deal beforehand, or even offering an opt-out deal. A spokesperson from Taylor & Francis claimed they sold to “help improve relevance and performance of AI systems,” while also claiming they were continuing to ensure authors were being paid for royalty payments in agreement with contracts [3]. Without widespread control, it won’t be the research institutions, authors, or historians dictating terms, but the other way around with AI companies and initiatives calling the shots.

A compilation of news headlines about Taylor & Francis Group's decision to sell academic research for AI training, featuring the Taylor & Francis logo at the centre of the image.

Usage of AI in Historical Research

Artificial intelligence technology can offer a variety of benefits to researchers including, but not limited to, “algorithmic search advancement;...metadata analysis; [ease of] cataloguing materials; the use of optical and intelligent character recognition and other software to digitize and make material accessible online; and the ability to do complex bibliometric and bibliographic analyses to trace research connections across disciplines in ways humans cannot recognize” [4]. Some may see AI technology and its benefits as a replacement for the role of human labour; but these tools do not come without risks and are usually difficult to pinpoint in an easily explainable manner. Data loss and breaches are a consistent risk, especially with the rising number of cyberattacks, competitive monopolies amongst data-sharers, and third-party organizations giving access to data without fair compensation, such as the Taylor & Francis case [5]. Many are also worried about the “outsourcing” of work, or the fact that AI could simply get it wrong [6].

While all these benefits and risks remain true, researchers across the world have harnessed these new technologies to aid them in their understanding and organizing of the past. The sheer amount of physical documents, photographs, and objects available in museums, libraries, archives, and other similar institutions is overwhelming to parse through, and often cataloguing information or metadata might not even exist for someone to be able to find what they’re looking for.

The Library of Congress in Washington D.C., a federal cultural institution and the world’s largest library, is currently undergoing a massive project with the aim of using artificial intelligence and machine learning to help cataloguers in the generation of metadata by using a backlog of uncatalogued ebooks [7]. In 2009, Johannes Preiser-Kapeller, a professor at the Austrian Academy of Sciences studying fourteenth-century Byzantine Church registries, built a database of hundreds of documents that would aid him in studying connections between individuals. His machine learning software aided him in discovering how the community's social network was maintained by the hidden contributions of women in their society. While he’s continued to work with his model in other projects, the model has had many limitations that require human intervention to understand it in greater context [8].

Researchers at the University of Antwerp developed an AI tool to help them in the recognition of script types of certain medieval manuscripts. Paleography, which is the study of ancient writings, and its related skills are arduous to learn, and dating and authenticating texts is extremely difficult and subjective. Online databases that are amassing digital reproductions of manuscripts with incomplete or outdated metadata also called on the need for something to be done [9]. In their conclusions, they found that while categorizing script into twelve different styles, the reality of it was much more complex and that the technology allowed them the opportunity to face their own assumptions about their research [10].

Companies that create and sell AI tools are now even beginning to advertise how their technology can assist in the study of history. Using examples of 19th-century Japanese photographs, Eden AI claims their technology can analyze faces and objects to detect image similarity, that it can analyze landmarks to detect artistic movements and specific time periods, and even detect “explicit or offensive material” for providing warnings to “protect vulnerable audiences.” Their goal is to sell a product that they believe could be of use to historians and researchers, and as far as their website shows, it doesn’t appear any historians worked on developing the software [11].

If you’ve been on TikTok or Instagram Reels in the last several weeks, you’ve likely seen a video with millions of views depicting a day in the life of a person living during a historical time period or event. These videos attempt to portray the daily life of a person from the moment they wake up, to the view outside their window, maybe even someone’s boils as they die from the Black Plague, or a Roman state building on fire from the eruption of Vesuvius in Pompeii in 79 CE. These videos are created with the effort in showing the common people that history isn’t just a thing in books. While the intent might be to inspire curiosity, in reality these videos are riddled with errors. 

Clips of “historical” AI videos from the TikTok account @histairy_films.

Since most AI models are trained with data from the last 10 to 15 years, computers have learned to only detect “modern” images using only the highest quality photos [12]. Some may think it doesn’t matter since these are videoes for social media, and although this argument does have its merits, they open a huge window to the general rise of misinformation. These videos could also be vehicals for spreading negative views about certain people and cultures, while opening doors for the creation of hateful or purposefully false videos [13]. 

Misinformation and AI

Misinformation, which is false or inaccurate information shared, has been an ongoing issue for many years and deserves an article, and many more, of its own. The risk of spreading misinformation and disinformation with AI always exists, as it’s extremely likely for these technologies to get it wrong. It’s healthy to have a bit of moderate skepticism towards something like new technology, and this includes AI tools. The sentiment that, “people [need] to view machine learning with critical detachment: as a useful tool, but one that’s fallible, just like people,” resonates with most examples of artificial intelligence technology usage within historical research [14].

Artificial intelligence and machine learning models can have many positive effects on historical research, aiding researchers in “filling-in-the-gaps” where information may be missing. They can help historians discover trends they might have missed themselves, or help them recontextualize certain aspects of their research. In almost all cases of AI usage, we can come to the conclusion that nothing can ever truly replace the human mind, especially when it comes to understanding our past. We are now only witnessing the beginning of the rise of AI tools and as our technology continues to develop and grow at remarkably fast rates, we’ll only see more. If anything, this technology proves how much human interference is still required, and AI likely won’t be taking our jobs anytime soon.

Endnotes

[1] Cole Stryker and Eda Kavlakoglu, “What Is Artificial Intelligence (AI)?,” IBM, February 14, 2025, https://www.ibm.com/think/topics/artificial-intelligence.

[2] Ibid.

[3] Matilda Battersby, “Academic Authors ‘shocked’ after Taylor & Francis Sells Access to Their Research to Microsoft AI,” The Bookseller, July 19, 2024, https://www.thebookseller.com/news/academic-authors-shocked-after-taylor--francis-sells-access-to-their-research-to-microsoft-ai.

[4] Matthew da Mota, “Libraries and Research Institutions Will Need Strong AI Governance,” Centre for International Governance Innovation, December 16, 2024, https://www.cigionline.org/articles/libraries-and-research-institutions-will-need-strong-ai-governance/.

[5] Ibid.

[6]  Moira Donovan, “How AI Is Helping Historians Better Understand Our Past,” MIT Technology Review, April 11, 2023, https://www.technologyreview.com/2023/04/11/1071104/ai-helping-historians-analyze-past/.

[7] Isabel Brador, “Could Artificial Intelligence Help Catalog Thousands of Digital Library Books? An Interview with Abigail Potter and Caroline Saccucci,” The Library of Congress: The Signal, November 19, 2024, https://blogs.loc.gov/thesignal/2024/11/could-artificial-intelligence-help-catalog-thousands-of-digital-library-books-an-interview-with-abigail-potter-and-caroline-saccucci/.

[8] Donovan, “How AI Is Helping Historians Better Understand Our Past.”

[9]  Mike Kestemont, Vincent Christlein, and Dominique Stutzmann, “Artificial Paleography: Computational Approaches to Identifying Script Types in Medieval Manuscripts,” Speculum 92, no. S1 (October 2, 2017): s86-87.

[10] Ibid, S109.

[11] “How Can AI Contribute to Art Historical Analysis and Research?,” Eden AI, accessed March 10, 2025, https://www.edenai.co/post/how-can-ai-contribute-to-art-historical-analysis-and-research.

[12]  Donovan, “How AI Is Helping Historians Better Understand Our Past.”

[13] Yasmin Rufo, “How Accurate Are the Viral Tik Tok AI POV Lab History Videos?,” BBC News, February 22, 2025, https://www.bbc.com/news/articles/cy87076pdw3o.

[14] Donovan, “How AI Is Helping Historians Better Understand Our Past.”

Bibliography

Battersby, Matilda. “Academic Authors ‘shocked’ after Taylor & Francis Sells Access to Their Research to Microsoft AI.” The Bookseller, July 19, 2024. https://www.thebookseller.com/news/academic-authors-shocked-after-taylor--francis-sells-access-to-their-research-to-microsoft-ai.

Brador, Isabel. “Could Artificial Intelligence Help Catalog Thousands of Digital Library Books? An Interview with Abigail Potter and Caroline Saccucci.” The Library of Congress: The Signal, November 19, 2024. https://blogs.loc.gov/thesignal/2024/11/could-artificial-intelligence-help-catalog-thousands-of-digital-library-books-an-interview-with-abigail-potter-and-caroline-saccucci/. 

Donovan, Moira. “How AI Is Helping Historians Better Understand Our Past.” MIT Technology Review, April 11, 2023. https://www.technologyreview.com/2023/04/11/1071104/ai-helping-historians-analyze-past/. 

“How Can AI Contribute to Art Historical Analysis and Research?” Eden AI. Accessed March 10, 2025. https://www.edenai.co/post/how-can-ai-contribute-to-art-historical-analysis-and-research

Kestemont, Mike, Vincent Christlein, and Dominique Stutzmann. “Artificial Paleography: Computational Approaches to Identifying Script Types in Medieval Manuscripts.” Speculum 92, no. S1 (October 2, 2017). https://doi.org/10.1086/694112. 

Mota, Matthew da. “Libraries and Research Institutions Will Need Strong AI Governance.” Centre for International Governance Innovation, December 16, 2024. https://www.cigionline.org/articles/libraries-and-research-institutions-will-need-strong-ai-governance/. 

Rufo, Yasmin. “How Accurate Are the Viral TikTok AI POV Lab History Videos?” BBC News, February 22, 2025. https://www.bbc.com/news/articles/cy87076pdw3o. 

Stryker, Cole, and Eda Kavlakoglu. “What Is Artificial Intelligence (AI)?” IBM, February 14, 2025. https://www.ibm.com/think/topics/artificial-intelligence.