AI’s potential to rediscover history is massive—but it presents challenges around bias and accuracy.

In the news

It’s tempting to think of history as a locked, fixed discipline. What happened, happened. It says so right there in your middle-school textbook.

The truth is far more complex. History is surprisingly malleable, shifting with intellectual trends, fresh discoveries—and, now, technologies like AI. “The application of modern computer science to the distant past helps draw connections across a broader swath of the historical record than would otherwise be possible,” historians say.

Applications of AI include mapping the relationships among 14^th-century bishops (and thus better understanding their relative influence); breaking codes in secret messages that had stymied scholars for 400 years; and creating three-dimensional reconstructions of razed neighborhoods.

The Cognizant take

“Historians are already using AI in various ways to enhance their research,” says Aakash Shirodkar, a Senior Director in Cognizant’s AI & Analytics Practice. “They analyze historical documents, images and artefacts; examine astronomical tables; decipher lost languages; identify patterns in past events; and restore or recreate archaeological artefacts.”

He points to a Stanford University research team that developed a machine learning algorithm to analyze the text of historical documents and identify the emotions expressed by the authors. Another example is a system developed at MIT's Computer Science and Artificial Intelligence Laboratory that can decipher a lost language without prior knowledge of its relationship to other languages.

“These are just a few examples of how historians have used AI to better understand our past and solve ancient puzzles,” Aakash says. But he’s quick to note concerns about potential biases and distortions introduced by these algorithms.

Therefore, when using AI in historical research, it's important to consider ethical considerations. One is the potential for AI-generated data to manipulate public opinion or create a false sense of security, Aakash says. Another issue is the risk of introducing bias or outright falsification into the historical record.

Moreover, nefarious intent isn’t a prerequisite; Aakash says poorly collected data can also reinforce existing gaps in the historical record. Finally, AI-driven historical research can be used for unethical purposes, such as manipulating data or drawing false conclusions.

To address these concerns, Aakash believes, historians, researchers and regulators “should prioritize transparency in training data, mandatory disclosures and appropriate labeling to allow adequate evaluation by historians.” It's essential to use the data responsibly and avoid manipulating public opinion or creating a false sense of security.

“History represents an almost unlimited swath of data waiting for AI to unlock its secrets,” Aakash says. “With the help of cloud computing and AI's ability to solve problems like image recognition, regeneration and restoration, as well as pattern recognition and text analysis, AI has a broad enough repertoire to help historians better understand our past and solve ancient puzzles quickly and efficiently.”