Artificial Intelligence: Is There Any Possible Application to History?


by Tommaso Costanzo, PhD candidate in Science of Advanced Materials.

Two of the most thought-provoking things of being married with someone studying a different discipline are the discovery of unexpected similarities and the possibility to learn from each other. For example, I am a chemist, and it was only while chatting with my wife, who studies history, that I came to realize that there are interesting similarities in our research methods, and that artificial intelligence (AI) could find useful applications not only in sciences, but also in the humanities.

My work as a material chemist is to search new materials with better properties compared to the ones already known. In theory this task can be easy accomplished by simply mixing numerous substances at different concentrations. However, since the combinations are infinite, this brute force approach is very inefficient (and potentially dangerous, you do not want to blow up by mistake!). In general, scientists rely on the existing knowledge (for example, the periodic table) to predict what will be a good candidate material, which is then synthesized and characterized to see if it is better or worse than the previous one.

This entire research process can also be accomplished by “machines,” a.k.a. computers. In fact, what is most commonly known as AI can do this exact process for us: the computer is trained with an already known set of data (e.g. many materials and their properties), and when the training is completed, the machine can recognize patterns in the given dataset, classify them in smaller groups, and also predict new materials.

Of course, when I understood how AI works and what it can do for my research, I was like a child receiving a new gift. However, even if I was aware of the potential of AI, I did not immediately realize that it can be something useful in other fields like, for example, history. This understanding came only while discussing with my wife about her research and work as an historian. Hearing her problems and reflections on the historical research and method made me notice the similarity between what historians and AI do. Indeed, historians generally search documents, traces, and any other sort of proof about the past. From this set of “data,” which is not necessarily ordered nor complete, they have to classify, order, and try to find pattern(s) in order to interpret and understand what happened in the past. So, it is possible to notice that the AI I use in my chemistry research accomplishes similar tasks to those that an historian has to do on his/her own.

Even though this is a very general discussion, which just aims at stimulating reflections, I suppose that historians will be able to benefit from the application of AI to their research. For example, AI has the potential to help deciphering and translating ancient texts. In fact, at the University of Alberta, a computing science professor used AI to advance the deciphering of the 15th century Voynich manuscript. Another possible application could involve the recognition and categorization of images. Also, AI could, for example, potentially help ancient historians filling in the missing parts of fragmentary documents with the most statistically probable text.

Notwithstanding these intriguing potential applications, there are indeed several hurdles to overcome. For example, for AI to function, it needs digital data. Archives and libraries have been digitalizing more and more documents (which ironically is already a process requiring an AI!), but it is not possible to digitalize everything. Furthermore, even though specific kinds of AI can offer predictions and interpretations, they cannot substitute the interpretation done by a professional historian.

Sciences and humanities have more in common than one would usually think. For this reason, we should discuss more and learn from each other.