Determining the historical significance of current events is extremely difficult, suggests new study
Machine learning tools can be useful for historians to analyse large volumes of data and minimize noise, suggests a new study.
How do we know if an event is historic? An event’s historical significance depends on how it affects subsequent events in the future. But predicting this can be difficult: what may seem historic now may be deemed trivial by future generations. New research suggests that, even with machine learning tools, determining historical significance is difficult but these tools can still help historians.
In the study, Joseph Risi and others tested if machine learning could predict the importance of an event in the future. To do this, they analyzed nearly two million declassified electronic cables from the United States Department of State between 1973 and 1979. At that time, these cables were the dominant form of communication between the US government and its embassies around the world. More importantly, these cables included information about both important events (such as the Iranian revolution) and trivial events (such as embassy social functions). This allowed the authors to score cables by their perceived importance at that time depending on several factors including whether it was designated for high-level attention.
Using a machine learning model, these important cables were then compared with another dataset of important cables compiled decades later by professional historians. Of the nearly two million cables, less than 1% were found in the historians’ dataset.
Interestingly, the cables perceived to be the most important at that time were only slightly more likely to be classified as important by historians later compared to cables perceived as less important. According to the authors, this suggests that historical significance is extremely difficult to predict. Despite this difficulty, the authors suggest that machine learning tools can still be useful for historians who can use them to analyze large volumes of information, minimize noise and focus on smaller sets of data.