This blog post presents an overview of the opportunities and challenges artificial intelligence (AI) presents for those engaging with history in a novel technological age.

Artificial intelligence (AI) is transforming industries worldwide, and the historical profession is no exception. Ever since the times of Herodotus and Thucydides, we historians have generally relied on tried and tested methods – archival research, textual analysis, and manual data collection – to deliver rigorous analyses of the past. We gather evidence about people, places, and things and analyze changes, causes, and effects. The arrival of AI, however, presents new opportunities and challenges to the ways we do history. Achievements that were once only possible for those of us with a university degree(s), a library card, and a lot of time on our hands are now readily available for anyone with an internet connection. Nevertheless, its over-reliance on automation and accusations of bias raise serious ethical and methodological challenges which we historians cannot ignore.

Without historical records to examine, the historical profession simply would not exist. With efforts to digitize these records picking up pace following the COVID-19 pandemic, AI processes have completely transformed the way in which historians can work with them. Digital archives, powered by machine-learning algorithms, allow vast collections of historical documents to be scanned, categorized, and made searchable. In these archives, AI-powered Optical Character Recognition (OCR) technology has evolved to transform scans of old manuscripts, newspapers, books, and other materials into machine-readable text. These algorithms can classify documents by date, topic, or author, significantly reducing the time needed for manual categorization. Resources from across the world in different languages and from different eras are now within the reach of everyone, from students of history to professional historians.

When historians access these records, we also look for patterns across time, space, and sources. Through the AI tool of text-mining, historians can trawl through volumes of materials at a previously incomprehensible pace and identify themes, sentiments, and trends in historical texts. Social historians studying the Industrial Revolution can track when language associated with technological advancements changed. Historians of religion – such as myself – can map out when terms such as “Buddhism” appeared and how its interpretation and meaning evolved. AI-powered geospatial analytical tools, such as Geographic Information Systems (GIS), further help environmental and urban historians to understand historical events through analysis of spatial relationships. As historians, we can therefore focus our energies on interpreting different data sets rather than seeking them out in the first place.

Furthermore, historians can now use AI to reconstruct historical environments and events far more effectively than the written word alone. Long gone are the days when articles, monographs, and conference presentations were the only ways in which scholars could bring to life the stories and people of ancient civilizations and societies. Using novel 3D technology, Swedish archaeologists and historians digitally reconstructed an entire block of houses at Pompeii to their original state before the 79 CE eruption of Mount Vesuvius. Relying on evolutions in mass spectrometry analysis, German researchers worked with art historians to conclusively date and colorize Afghanistan’s destroyed Bamiyan Buddhas. Predictive algorithms have further enabled the development of deep neural networks like “Ithaca” which can restore damaged texts and date ancient documents with high levels of accuracy.

Most importantly, however, AI products are making it easier for amateur scholars, educators, and members of the public to engage with history in ways that were not previously possible. Natural Language Processing (NLP) tools, such as the Google Books Ngram Viewer, can analyze the frequency of words and phrases in millions of digitized books, allowing us all to trace cultural and linguistic trends over time. The “SlaveVoyages” database provides another example of AI educating the public through its descriptions of the transatlantic slave trade, allowing those without prior historical training to learn about its causes and effects. Through virtual assistants or techniques such as Augmented Reality (AR), as in the British Museum’s “The Living Museum” project and the Louvre’s “Mona Lisa: Behind the Glass” experience, AI can also create more engaging exhibits to broaden participation in historical discourse.

With all these innovations and new tools, what then will become of the historian? As any good history student will tell you, historical data often reflects the biases of its time. Following alarming examples of leading chatbots providing historically inaccurate responses to certain prompts about the Holocaust, a recent UNESCO report called on tech companies to draft ethical guidelines for AI development and use to prevent malignant actors from using AI to spread hate. AI-generated images of events such as the bombing of Hiroshima and the assassination of President Kennedy also demonstrate the future risk of fake visuals being presented as fact, which risks weakening public trust in visual materials as a reliable resource for examining the past. Even Google, too, was forced to apologize after its Gemini AI tool depicted historical figures such as the US Founding Fathers and Nazi combatants as people of colour – remarkably as an overcorrection to the problems of racial bias in AI products.

Moreover, while AI can certainly process data, it lacks the nuanced understanding and critical thinking that human historians bring. For these AI technologies, “doing history” means trawling through mainly online datasets and producing responses that are presented to the user in the most statistically common form, with results that trend towards conformity and uniformity. While producing history certainly requires amassing vast amounts of information and recognising patterns, these are only two facets of the historian’s wider skillset. It is not enough to read a lot, work out the established narrative, and then map out the different ways to present an argument. Historians must also do something entirely different with the information they gather. They take it, break it down piece by piece, and start again. They place it into different contexts. They use it to reframe or disprove what everyone currently believes.

AI alone cannot and does not function in this way.

The integration of AI into the historical profession is, of course, still in its infancy. Indeed, emerging technologies promise to further revolutionize the field. AI has changed – and is continuing to change the world – but will always require mediation in some form or another. Notwithstanding the risks and dangers associated with its emergence, it holds remarkable potential to transform the practice of history for the better. So, will the proliferation of AI products and solutions spell the end for the historical profession? Will AI really replace our history teachers and professors? Probably not.

<figure class="wp-block-image size-full is-resized"><figcaption class="wp-element-caption">Bhadrajee Hewage recently completed his DPhil in History at the University of Oxford, researching trends in subcontinental Buddhism during the late colonial and early postcolonial periods. He also serves as a Graduate Outreach Tutor for Oxford’s Faculty of History.</figcaption></figure>

Cover image: Wikicommons, https://commons.wikimedia.org/wiki/File:Artificial-Intelligence.jpg

The post Histor(AI): Doing History in an AI World appeared first on On History.