This is the transcript of the talk I gave at the Tech4Good event I'm at this weekend in Albuquerque, New Mexico. The complete slide deck is here.
I want to talk a little bit about a problem I see – or rather, a problem I see in the “solutions” that some scientists and technologists and engineers seem to gravitate towards. So I want to talk to you about pigeons, operant conditioning, and social control, which I recognize is a bit of a strange and academic title. I toyed with some others:
I spent last week at the Harvard University archives, going through the papers of Professor B. F. Skinner, arguably one of the most important psychologists of the twentieth century. (The other, of course, being Sigmund Freud.)
I don’t know how familiar this group is with Skinner – he’s certainly a name that those working in educational psychology have heard of. I’d make a joke here about software engineers having no background in the humanities or social sciences but I hear Mark Zuckerberg was actually a psych major at Harvard. (So that’s the joke.)
I actually want to make the case this morning that Skinner’s work – behavioral psychology in particular – has had profound influence on the development of computer science, particularly when it comes to the ways in which “programming” has become a kind of social engineering. I’m not sure this lineage is always explicitly considered – like I said, there’s that limited background in or appreciation for history thing your field seems to have got going on.
B. F. Skinner was a behaviorist. Indeed, almost all the American psychologists in the early twentieth century were. Unlike Freud, who was concerned with the subconscious mind, behaviorists like Skinner were interested in – well, as the name suggests – behaviors. Observable behaviors. Behaviors that could be conditioned or controlled.
Skinner’s early work was with animals. As a graduate student at Harvard, he devised the operant conditioning chamber – better known as the Skinner box – that was used to study animal behavior. The chamber provided some sort of response mechanism that the animal would be trained to use, typically by rewarding the animal with food.
During World War II, Skinner worked on a program called Project Pigeon – also known as Project Orcon, short for Organic Control – an experimental project to create pigeon-guided missiles.
The pigeons were trained by Skinner to peck at a target, and they rewarded with food when they completed the task correctly. Skinner designed a missile that carried pigeons which could see the target through the windows. The pigeons would peck at the target; the pecking in turn would control the missile’s tail fins, keeping it on course, via a metal conductor connected to the birds’ beak, transmitting the force of the pecking to the missile’s guidance system. The pigeons’ accuracy, according to Skinner’s preliminary tests: nearly perfect.
As part of their training, Skinner also tested the tenacity of the pigeons – testing their psychological fitness, if you will, for battle. He fired a pistol next to their heads to see if loud noise would disrupt their pecking. He put the pigeons in a pressure chamber, setting the altitude at 10,000 feet. The pigeons were whirled around in a centrifuge meant to simulate massive G forces; they were exposed to bright flashes meant to simulate shell bursts. The pigeons kept pecking. They had been trained, conditioned to do so.
The military canceled and revived Project Pigeon a couple of times, but Skinner’s ideas were never used in combat. “Our problem,” Skinner admitted, “was no one would take us seriously.” And by 1953, the military had devised an electronic system for missile guidance, so animal-guided systems were no longer necessary (if they ever were).
This research was all classified, and when the American public were introduced to Skinner’s well-trained pigeons in the 1950s, there was no reference to their proposed war-time duties. Rather, the media talked about his pigeons that could play ping-pong and piano.
Admittedly, part of my interest in Skinner’s papers at Harvard involved finding more about his research on pigeons. I use the pigeons as a visual metaphor throughout my work. And I could talk to you for an hour, easily, about the birds – indeed, I have given a keynote like that before. But I’m writing a book on the history of education technology, and B. F. Skinner is probably the name best known with “teaching machines” – that is, programmed instruction (pre-computer).
Skinner’s work on educational technology – on teaching and learning with machines – is connected directly, explicitly to his work with animals. Hence my usage of the pigeon imagery. Skinner believed that there was not enough (if any) of the right kind of behavior modification undertaken in schools. He pointed that that students are punished when they do something wrong – that’s the behavioral reinforcement that they receive: aversion. But students are rarely rewarded when they do something right. And again, this isn’t simply about “classroom behavior” – the kind of thing you get a grade for “good citizenship” on (not talking in class or cutting in the lunch line). Learning, to Skinner, was a behavior – and a behavior that needed what he called “contingencies of reinforcement.” These should be positive. They should minimize the chances of doing something wrong – getting the wrong answer, for example. (That’s why Skinner didn’t like multiple choice tests.) The reinforcement should be immediate.
Skinner designed a teaching machine that he said would do all these things – allow the student to move at her own pace through the material. The student would know instantaneously if she had the answer right. (The reward was getting to move on to the next exciting question or concept.) And you can hear all this echoed in today’s education technology designers and developers and school reformers – from Sal Khan and Khan Academy to US Secretary of Education Betsy DeVos. It’s called “personalized learning.” But it’s essentially pigeon training with a snazzier interface.
“Once we have arranged the particular type of consequence called a reinforcement,” Skinner wrote in 1954 in “The Science of Learning and the Art of Teaching,” "our techniques permit us to shape the behavior of an organism almost at will. It has become a routine exercise to demonstrate this in classes in elementary psychology by conditioning such an organism as a pigeon.”
“…Such an organism as a pigeon.” We often speak of “lab rats” as shorthand for the animals used in scientific experiments. We use the phrase too to describe people who work in labs, who are completely absorbed in performing their tasks again and again and again. In education and in education technology, students are also the subjects of experimentation and conditioning. In Skinner’s framework, they are not “lab rats”; they are pigeons. As he wrote,
…Comparable results have been obtained with pigeons, rats, dogs, monkeys, human children… and psychotic subjects. In spite of great phylogenetic differences, all these organisms show amazingly similar properties of the learning process. It should be emphasized that this has been achieved by analyzing the effects of reinforcement and by designing techniques that manipulate reinforcement with considerable precision. Only in this way can the behavior of the individual be brought under such precise control.
If we do not bring students’ behavior under control, Skinner cautioned, we will find ourselves “losing our pigeon.” The animal will be beyond our control.
Like I said, I’m writing a book. So I can talk at great length about Skinner and teaching machines. But I want folks to consider how behaviorism hasn’t just found its way into education reform or education technology. Indeed, Skinner and many others envisioned that application of operant conditioning outside of the laboratory, outside of the classroom – the usage (past and present) of behavior modification for social engineering is at the heart of a lot of “fixes” that people think they’re doing “for the sake of the children,” or “for the good of the country,” or “to make the world a better place.”
Among the discoveries I made – new to me, not new to the world, to be clear: in the mid–1960s, B. F. Skinner was contacted by the Joseph P. Kennedy Jr. Foundation, a non-profit that funded various institutions and research projects that dealt with mental disabilities. Eunice Kennedy Shriver was apparently interested in his work on operant behavior and child-rearing, and her husband Sargent Shriver who’d been appointed by President Johnson to head the newly formed Office of Economic Opportunity was also keen to find ways to use operant conditioning as part of the War on Poverty.
There was a meeting. Skinner filed a report. But as he wrote in his autobiography, nothing came of it. “A year later,” he added, “one of Shriver’s aides came to see me about motivating the peasants in Venezuela.”
Motivating pigeons or poor people or peasants (or motivating peasants and poor people as pigeons) – it’s all offered, quite earnestly no doubt – as the ways in which science and scientific management will make the world better.
But if nothing else, the application of behavior modification to poverty implies that this is a psychological problem and not a structural one. Focus on the individual and their “mindset” – to use the language that education technology and educational psychology folks invoke these days – not on the larger, societal problems.
I recognize, of course, that you can say “it’s for their own good” – but it involves a great deal of hubris (and often historical and cultural ignorance, quite frankly) to assume that you know what “their own good” actually entails.
You’ll sometimes hear that B. F. Skinner’s theories are no longer in fashion – the behaviorist elements of psychology have given way to the cognitive turn. And with or without developments in cognitive and neuroscience, Skinner’s star had certainly lost some of its luster towards the end of his career, particularly, as many like to tell the story, after Noam Chomsky penned a brutal review of his book Beyond Freedom and Dignity in the December 1971 issue of The New York Review of Books. In the book, Skinner argues that our ideas of freedom and free will and human dignity stand in the way of a behavioral science that can better organize and optimize society.
“Skinner’s science of human behavior, being quite vacuous, is as congenial to the libertarian as to the fascist,” writes Chomsky, adding that “there is nothing in Skinner’s approach that is incompatible with a police state in which rigid laws are enforced by people who are themselves subject to them and the threat of dire punishment hangs over all.”
Skinner argues in Beyond Freedom and Dignity that the goal of behavioral technologies should be to “design a world in which behavior likely to be punished seldom or never occurs” – a world of “automatic goodness.“ We should not be concerned with freedom, Skinner argues – that’s simply mysticism. We should pursue ”effectiveness of techniques of control“ which will ”make the world safer." Or make the world totalitarian, as Chomsky points out.
Building behavioral technologies is, of course, what many computer scientists now do (perhaps what some of you do cough FitBit) – most, I’d say, firmly believing that they’re also building a world of “automatic goodness.” “Persuasive technologies,” as Stanford professor B. J. Fogg calls it. And in true Silicon Valley fashion, Fogg erases the long history of behavioral psychology in doing so: “the earliest signs of persuasive technology appeared in the 1970s and 1980s when a few computing systems were designed to promote health and increase workplace productivity,” he writes in his textbook. His students at his Behavioral Design Lab at Stanford have included Mike Krieger, the co-founder of Instagram, and Tristan Harris, a former Googler, founder of the Center for Humane Technology, and best known figure in what I call the “tech regrets industry” – he’s into “ethical” persuasive technologies now, you see.
Behavior modification. Behavioral conditioning. Behavioral design. Gamification. Operant conditioning. All practices and products and machines that are perhaps so ubiquitous in technology that we don’t see them – we just feel the hook and the urge for the features that reward us for behaving like those Project Pigeon birds pecking away at their target – not really aware of why there’s a war or what’s at stake or that we’re going to suffer and die if this missile runs its course. But nobody asked the pigeons. And even with the best of intentions for pigeons – promising pigeons an end to poverty and illiteracy, nobody asked the pigeons. Folks just assumed that because the smart men at Harvard (or Stanford or Silicon Valley or the US government) were on it, that it was surely right “fix.”