Did AI show that we are “a peaceful species” triggered by religion?No, but this episode shows how science media sometimes help mislead the public
There’s an old saying, “Don’t mistake the map for the territory.” That is, when reading a map, recognize that the map is an abstraction and an imperfect one at that. The map won’t show every feature. In fact, maps may intentionally ignore certain features so as to show others more clearly. For example, a street map leaves out the cars parked in driveways, even those that have been there so long that they seem to be permanent fixtures.
Below is a view of my neighborhood from Google Maps. Suppose someone looked at it and said, “That’s a strange place you live in—the streets are white and perfectly straight and the yards and houses are all grey.” Clearly, that person does not understand what a map is: a symbolic representation of features selected for a purpose. It is not a visual representation of what we would see.
Do you think that no one would make such a mistake? The mistakes commonly made in the professional science media can be even more egregious.
A recent study, “A Generative Model of the Mutual Escalation of Anxiety Between Religious Groups” (open access), provides a handy illustration. The researchers attempted to represent the escalation of anxiety between groups using a simple model. The resulting study was, at best, mildly interesting. The authors claim that they can model how mild events that increase anxiety can eventually lead to an overall average anxiety level that could exceed a threshold.
Bored yet? You should be.
So what did the professional science media have to say about this paper? In “Can artificial intelligence help stop religious violence?”, Mark Easton tells us at the BBC that the paper “indicates people are a peaceful species by nature.” Did the paper actually say that? Absolutely not—the paper simply assumed it. That’s like referring to the yards in the picture above as grey. The yards are grey because the map doesn’t offer color information!
The BBC wasn’t the only one. ScienceDaily also reports “The findings reveal that people are a peaceful species by nature.” Again, the findings say absolutely no such thing. But the ScienceDaily story goes further: “The study is built around the question of whether people are naturally violent, or if factors such as religion can cause xenophobic tension and anxiety between different groups, that may or may not lead to violence?” As we will see, this characterization of the study is completely bogus.
In fact, nearly every claim about the paper seems to misunderstand how computer models work generally and how they worked in this paper in particular. First, there is nothing particularly “religious” about the criteria used in the model. In computer models, you can name the pieces of the model however you wish. The authors of the software simply happened to assign religious names to the components of the model. There was hardly anything religious about it apart from that.
According to the BBC article, the study shows that “The most risky situations are when the difference in the size of two different religious groups is similar and people encounter ‘out-group members’ more regularly, perceiving them as dangerous.”
Did the study show that? Let’s look at it:
The goal of our model is to generate mutually escalating xenophobic anxiety between two religious groups under theoretically sound conditions that are consistent with TMT, SIT, and IFT.
So the goal of the model is not to mimic human psychology generally and see what happens but to actually to make the effect happen, using elements from previous theories. The authors go on to say,
We chose the main elements of the architecture (agent traits, networks, hazards, group identities, etc.) with the goals of the model in mind. While many other variables and factors are relevant for mapping a phenomenon as complex as religious conflict, we aimed for a level of abstraction that captured the most salient and empirically-researched mechanisms that bear on this type of intergroup conflict.
There was no possibility that anxiety levels that could be precursors to social violence would not occur. The programmers would tweak their program until they did. There was no possibility that the study would discover some other cause of the impending violence because the only allowed possibilities were group identities.
Interestingly, the model studied didn’t even address violence directly. It featured a built-in marker for anxiety. The authors checked to see if any part of the system passed a threshold number, on the theory that if the number got really high, violence might follow.
The study did not assert that humans are normally peaceful, because neither peace nor violence was modeled by the simulation. The programmers set a number that they labeled “anxiety” to start off at a low level. They showed that their code successfully increased that number past a threshold.
So why are two different outlets reporting something that the paper doesn’t say? ScienceDaily offers a clue: “Materials provided by University of Oxford. Note: Content may be edited for style and length.” It would be less disturbing if such misunderstandings came from reporters. But it seems that they originated from one of the universities affiliated with the study. Where would Oxford get its information? Probably from the researchers themselves.
Unfortunately, most of the public knows about science only through science media professionals. And it is apparent that science media professionals often know little to nothing of what they are talking about. They can’t distinguish the map, the territory, the significance, or even the difference between inputs and outputs. Even more unfortunately, based on the fact that these articles were based on materials provided by the universities, this misinformation may be coming directly from the researchers.
Note: The paper is by LeRon Shultsa, Ross Goreb, Wesley J. Wildmanc, Christopher Lynchb, Justin E. Laned and Monica Toft. A Generative Model of the Mutual Escalation of Anxiety Between Religious Groups. The Journal for Artificial Societies and Social Stimulation, 2018 DOI: 10.18564/jasss.3840
Jonathan Bartlett is the Research and Education Director of the Blyth Institute.
Also by Jonathan Bartlett: Self-driving vehicles are just around the corner On the other side of a vast chasm…
Guess what?: You already own a self-driving car. Tech hype hits the stratosphere