There are many rules, both written and unwritten, to modern academic work. Perhaps the best (if depressing) recent explanation of unwritten rules is Daniel Nexon’s blunt warning about the necessity of self-promotion for academic work. Everyone is pretty much caught in the REF/accessibility trap (to wit: “Publish in a format the general public can’t afford, or perish”). As an early career researcher, not paying attention to these structural issues is career suicide. To this, I might add that I have a new book on American targeted killings and transnational war available from Hurst & Co. in the UK and Oxford University Press in the United States. Ahem. But I digress. The reason I’m writing this is to pay homage to perhaps the most flagrant piece of unwritten-rules-of-modern-academia rule-breaking that I’ve come across in recent years: Rainer Hegselmann’s 155 page article in the Journal of Artificial Societies and Social Simulation.
The purpose of this piece is to convince you (likely an academic, or someone interested in academic work) to read a 155 page piece of work on the history of computational social science. Depending on how much you care about CSS, this might be a big ask, but hear me out. Hegselmann’s article (which, in book form, would be something like a novella, which reminds me that we don’t seem to have an equivalent term for the gap between article and book in academic knowledge production) is as much a meditation on multiple themes (power, race, fame, technology and communication) as it is a work of intellectual history.
The short version of the story is this: James Minoru Sakoda was a pioneer of computational social science, yet a man who was more remembered for his origami than his academic output at the end of his life. Thomas C. Schelling (aka that Schelling aka you’ve definitely read him at some point if you do social science) was perhaps one of the most celebrated academics of the 20th Century. Their careers intersected, in a sense, in 1971, when first Sakoda, then Schelling, published articles in The Journal of Mathematical Sociology. Sakoda’s article, published first, used a model that was “much more general and flexible than Schelling’s” but it was Schelling’s model, described in his article, published second, that got all the credit. Schelling goes down as the inventor of a famous model that can model processes of segregation from simple rules of neighbourhood preference (eg how an individual agent perceives their immediate surroundings), Sakoda, whom Hegselmann identifies as the very first pioneer of computational social science, is mostly remembered by origami enthusiasts.
In academic terms: What gives? Hegselmann, who spent 20 years researching the article, is quick to discount malfeasance:
My story is somewhat thrilling (at least the process of understanding, what has happened was it for me). But it is not a thriller! No crime happened, no conspiracy was involved. No discrimination whatsoever was at work. As to the main actors, all rules of honest scientific work, citation, and giving credit were abided. By all standards, nothing “unethical” is part of the story. Something went seriously wrong, but, as to the main actors, nobody did anything wrong. In retrospect, what happened was due to an interplay of fairly simple factors and mechanisms.
As it turns out, that “interplay of fairly simple factors and mechanisms” is both enlightening and thought provoking. Although the connection between race and segregation models is clear to anyone that encounters them (segregation models demonstrate the collective behaviour of stark self-segregation arising from relatively small individual preferences, although segregation in practice often involves many other factors, like red-lining etc), Hegselmann’s research into Sakoda’s life story points to another intellectual tie —the mass internment of ethnic Japanese U.S. citizens during World War 2. And so begins a dozen or so pages that tie the origin of agent-based modelling to social science research programmes on interned U.S. citizens, and the webs of funding and publications tied to them. Sakoda’s dissertation arising from this research isn’t published due to a number of issues. I can’t accurately summarise them here, but if you are at all interested in the relationship between researcher, subject, and funders (you should be!) then it’s worth reading this section in full. Reading this early phase of Sakoda’s life, one also gets the sense of missed opportunities each way — the Department of Defense, according to Sakoda, refused to hire him in the early 1950s because of racism — Sakoda went on to pioneer computational approaches in academia, but that single damning line reads like a loss for America as a whole.
If you’re not interested in agent-based modelling, or the history of social science, the inversion of Sakoda and Schelling’s reputations should still serve as a reason to read Hegelsmann’s article. Here, the article deals with core issues of originality, competence, and importance. How come a technically advanced, general, and adaptable model described by the pioneer of a field lost out to a simple model designed by an intellectual interloper that could be done with coins and a 3x3 grid? As academics, we’re taught that this is the wrong outcome, perhaps privately, we see lesser examples everywhere. The success of Schelling’s model was precisely that it could be done by anyone — that you didn’t, like Sakoda, need to understand how computers worked in order to do segregation modelling — and, perhaps as importantly, that it was a visual process. People could understand Schelling because they had to flip coins by hand, and this allowed them to get their heads around the processes at work. Sakoda’s model, by contrast, required computation, but more importantly, computation that was not representable to the people performing the modelling. Unlike today, where anyone interested in agent-based modelling can whip up a step-by-step visual model as they learn, the computers in Sakoda’s time weren’t capable of that. Moreover, it was impossible to disseminate Sakoda’s model because computers weren’t portable like they are today. Sakoda’s work was so advanced that it could be effectively demonstrated in research labs only, whereas Schelling’s could be sketched on a napkin. This is a very different environment to today. Hegelsmann notes that no other social scientists picked up Sakoda’s work in the 1970s, and at that point, most of them were interested in computational statistics, not computational modelling. Schelling’s model, in contrast, was quickly picked up by economists, and unlike Sakoda, Schelling built upon his original work. When advances in computation capability in the 1980s enabled a new generation of researchers to access computer resources and investigate computational approaches to existing problems, they reached for the Schelling that they had read and understood, and not Sakoda:
Thus, as to Sakoda’s model, all things considered, a sad and simple diagnosis suggests itself: when Sakoda’s time had come, when the skills and technical equipment that were necessary to realize his project, were really there, his research program was already forgotten. Sakoda’s model was simply not known and did not get known to the new generation that now was well equipped to start the research on it. Sakoda had published his article ten to fifteen years too early. Maybe that life punishes those that are late. But sometimes it punishes those that are early as well.
Hegselmann continues his analysis with a discussion of the Matthew-Effect, where important people get disproportionately more credit than the less important. By the time Schelling was looking at this kind of modelling, he was already a world-famous academic and policy professional. I think this section of the article has added resonance in today’s world — particularly amongst people whose research covers technical subjects that don’t get purchase with the significant policy circles. Unfortunately I’m still waiting for Daniel Drezner’s The Ideas Industry to drop through my letterbox, but this article seems to approach similar themes: how do the people that academics connect and network with shape the formation, reception, and propagation of their ideas? Hegelsmann’s study underlines their importance, and his final section gives plenty of food for thought.
I’ll finish there, because no-one wants to read an over-long blog post about an article that could be a book. But I’ll say this: despite all the doom and gloom about the future of the academy, there’s still room for 20-year pet projects, and electrifying pieces of research. As an early career researcher, that’s heartening, to say the least.