History by Algorithms

What was the structure of Jerusalem's social network in 70 CE? How were rumors spread, and how quickly did they reach their destination? How many nobles were jailed at the Jericho hippodrome, threatened to be executed to ensure that on the day of the Herod death, there would be great sorrow instead of joy? And how many people were invited to the wedding of hottest celebrity, the High Priest? "Social networks in the past, just like in the present, follow rules, and by observing a small set of guidelines we can construct historical social networks and study historical events such as the fall of the Second Temple from a sociological aspect," explains Prof. Zvika Lotker. He details these guidelines in the sixth chapter of his latest book, History by Algorithms, dedicated to the mathematics of history and how computers, and AI in particular, will change the way we research, study, and grasp history. "Marshall McLuhan said that the medium is the message – but the medium has changed. We are in the midst of the AI revolution, and nearly every aspect of our life is going to change, history included," he asserts.

History shapes who we are

Prof. Lotker says that he has always loved history. "As a child, I would chat with my dad about history and spent years reading and studying it as a hobby. It was great fun to explore this link between history and computer science," he shares. "But from the moment I started writing this book, it completely changed the way I look at history and the process of manufacturing it. We're used to think about history as a realm of academics who sit down and research and write, but if you really think about it, we're all historians. We live in this world, consume news, experience the changing world around us, and eventually have to tell ourselves our own life story."

How many soldiers were killed on the third day of the Battle of Thermopylae?

The book is divided into three sections. Section one is dedicated to the mathematics of history and offers mathematical tools to depict and study history. "Machine learning and deep learning tools incorporate architectures and neural networks that know how to tackle these tasks. If a machine needs to identify a picture of a cat, for example, before it knows that it must locate a creature with four legs and a tail, it has to know what a pixel is, what color is, how to clear up noise; these are the foundations that the network uses. Therefore, the first section of the book is dedicated to mathematical foundations: the magnitudes we refer to, categorization of history, or graph theory," says Prof. Lotker. "Looking at history through this lens reveals fascinating phenomena. For instance, if we look at how global money increased over the course of history, yet accumulated by a smaller number of people, we can conclude that mathematically speaking, the rich are not a linear part of society."

One of the chapters in the first section is dedicated to a mathematical observation of the history of wars. "History's most famous battles can be translated into a set of differential equations that is temporally non-homogenous, and receive a practically live historical depiction. Battles have two phases: first they face each other, then one of the armies rushes forward or flanks the other. That is how most of history's greatest battles played out. Mathematics can provide plenty of parameters that are absent from historical descriptions, even how long a battle lasted or how many soldiers were injured or killed in each phase."

Every hero has a villain

The second section of the book explores how the mathematics of history can be applied. For instance, how mathematics helps provide rules to history. "In most cases, the history we know is that of the core, the ruling class; we often wonder if it also applies to the common person. In the book's eighth chapter, I formulate a very simplistic model that shows how, under certain conditions, the answer is yes: we can trust the history of the core to depict the history of the periphery," explains Prof. Lotker. To the same end, we can create a model that depicts a society undergoing a revolution – say, the Roman Empire going from a pagan to a Christian society – and try to reach conclusions. The laws of physics say that revolutions precede the ideas of the majority of the population and happen much faster than we tend to think. If we apply that to the AI revolution, we can deduce that it will happen far quicker than we can realize or imagine."

Another aspect explored in the book is that of historic memory, the formation of collective memory, how a narrative is shaped when witnesses are still alive – for example, the memory of the Holocaust or October 7^th – and what happens to memories after the death of the witness. "After a certain point there are no more living witnesses, and all that remains is the written text; it will determine the history of the event," Prof. Lotker says. "Mathematics shows how every historical figure known as 'The Great' had an anti-hero, a historical figure shaped as the villain, and that this narrative was shaped long enough after the event's occurrence so that there would be no witnesses to claim 'wait, I remember and that's not how it happened'. The villain of Elizabeth I, aka Elizabeth the Great, was Richard III who ruled some 70 years before her. Roman emperor Augustus had Neron, who ruled some 70 years after him. Before or after, it doesn't matter in the perspective of time. How this text was written and shaped is no longer reality – it is how we grasp reality and what story we choose to tell."

Pop quiz: Are you a fascist?

The third section of the book is about the future and how AI will change the study, documentation, and analysis of historical texts. "AI allows us to read, process, and catalog insane quantities of information, and we can use methods of machine learning to analyze this information, whether it is written or filmed," clarifies Prof. Lotker. "For example, I show how we can use AI to analyze the body language of politicians. I trained the machine using the film The Great Dictator by Charlie Chaplin, which offers an artistic discussion of fascism. I took two scenes, one of his speech as a dictator, the other of his speech as a democrat, and let the machine analyze it and learn to discern between fascist and liberal body language. I then let the machine analyze the first debate between Trump and Clinton in 2016. Not surprisingly, the machine concluded that Donald Trump employs the body language of a fascist. I take the liberty to predict the future and suggest that, a few years from now, AI analysis of body language in debates and political speeches will be standard practice."

In tandem with writing his book, Prof. Lotker has developed a groundbreaking course: Computational History. "It is the first course of its kind in the world, where we take digital documents of events, see what computational tools from the forefront of technology can process historical data, and generate history." The course will be offered to students of computer and data engineering starting next semester, "and I believe it will be a good fit for humanities students, because it is part of their future."

The book History by Algorithms was published by Springer and can be downloaded from the publisher's website or the university's website.