Interesting study analysing the relationship between internal vs instrumental (the author’s take on intrinsic vs extrinsic) motivation as revealed in entry questionnaires for West Point cadets and long-term success in army careers. As you might expect, those with intrinsic motivation significantly outperformed those with extrinsic motivation on every measure.

What is particularly interesting, however, is that extrinsic motivation crowded out the intrinsic in those with mixed motivations. Having both extrinsic and intrinsic motivation is no better than having extrinsic motivation on its own, which is to say it is virtually useless. In other words, as we already know from hundreds of experiments and studies over shorter periods but herein demonstrated for periods of over a decade, extrinsic motivation kills intrinsic motivation. This is further proof that the use of rewards (like grades, performance-related pay, and service awards) in the hope that they will motivate people is an incredibly dumb idea because they actively demotivate. 

Address of the bookmark: http://m.pnas.org/content/111/30/10990.full

A dark tale for the halloween season

It was a dark and stormy night when Phil deMuth, an investment advisor, sat down to pen this article for Forbes. His voodoo incantations would raise from the dead the ghastly zombified remains of behaviourist dogmatist and appropriately named Skinner and, together, their nightmarish vision would turn the nation’s young into mindless zombies: apathetic, disenfranchised, undead and unthinking fodder suited only to sustaining the ghost of an industrial past. It is because of idiotic but superficially plausible ideas like the ones in this skillfully written article that I have to struggle to unteach, to try and often fail to help students to learn how to love learning again after years of having it beaten out of them. So, in case anyone is persuaded by the slick but outrageously wrong arguments of the article, or is one of the many educators that actually make use of this claptrap, this post is meant as a small antidote to the zombie plague.

A good beginning

DeMuth’s article starts very well. The first six paragraphs of the article present a fine and impassioned analysis of the failings of popular educational technologies and methods that strikes a well-aimed blow at the heart of many of the problems in existing educational systems, the atrociousness of traditional methods, and the unwitting replication of harmful and outmoded ways of teaching in poorly designed MOOCs and misuse of Khan tutorials (he actually attacks Khan tutorials themselves but, had he thought to check out a few rather than lump them with the rest, he would have discovered that they actually closely conform to his vision). deMuth might also have attacked much of university e-learning too on the same grounds. I could not agree more. DeMuth is absolutely right to bemoan the crazy transmission model underlying much of education and the appallingness of monolithic and intimidating exams and tests, not to mention the foolishness of replicating old and weak methods in new and shiny tools. His motives are unimpeachable and he makes a very strong, eloquently argued case. His solution though is not to change that system but to make it do the (wrong) job more efficiently. Here I take issue.

And then the horror starts

The remainder of the article is entertainingly and slickly written but the worse for that, because it carries a very dangerous message indeed.  In brief, it spends a while self-referentially demonstrating the value of programmed learning, then winds up by asking for programmers/behaviourist psychologists to produce modern equivalents of Skinner’s Teaching Machine (see illustration) so that the transmission model can work better and kids can pass more tests.

Behaviourism revisited

I’m amazed that anyone still thinks radical behaviourism, as espoused by Skinner, has any value whatsoever. I guess that some people learned this stuff before it was soundly discredited or, as I’m guessing was the case for deMuth, discovered it in passing without looking into what the rest of the world thought about it. Alas, such beliefs still do still persist. Indeed, we see behaviourist shortcuts all too regularly in education and industry to this day, even though hardly anyone who has followed any of the research over the past 40 years or so would find it at all acceptable. 

For those unfamiliar with Skinner’s radical behaviourist model, in brief, it was meant to apply a reductionist scientific method to discovering how animals (including people) learn. Recognizing that internal cognitive processes are hard to observe (and, in Skinner’s radical version of the theory, are themselves simply a consequence of external conditioning), the ‘behaviour’ part of the name is a reflection of that fact that this is what behaviourists concentrated on and, in Skinner’s case, the only thing that counted. Skinner only allowed for stimuli and responses that could be observed and measured – nothing else mattered. The brain for Skinnerian behaviourists was a black box about which they needed to know nothing apart from the effects of particular inputs and the observable outputs. They performed interventions and observed their effects on behaviour. Based on these observations (not uncommonly starting and sometimes ending with experiments on animals) those who tried to apply these methods in teaching sought to work out how to teach better, without ever having to make any assumptions about what was going on inside people’s minds. It’s a laudable goal, if Quixotic and utterly misguided. One big trouble with it (though far from the only one) is that it ignores our minds’ own inputs, that are often a great deal more significant than any external stimuli and that always modify them, in unpredictable ways, with complex effects that often fail to emerge until long after the stimuli have gone. We now know that reductionist methods simply don’t work in this context. Skinner lacked the framework of complexity and chaos theories that demonstrate the theoretical and practical impossibility of predicting even such simple causes and effects as the motion of a double pendulum, so it is perhaps a little forgiveable that he remained lost in a reductionist paradigm. We also now know that the operant conditioning methods that Skinner espoused are relatively ineffective in the short term and highly ineffective in the long term, so behaviourism fails to achieve much even on its own terms. Again, Skinner could not really be blamed for misunderstanding the significance of his results or their long-term weaknesses because such research was in its infancy while he was still alive.

So why did anyone ever believe in this stuff?

To a limited extent, behaviourism works. Among radical behaviourist ‘discoveries’, in large part guided by experiments in which Skinner was able to train animals like pigeons little by little to perform complex tasks, is the one focused on in this article: that small chunked lessons, with immediate feedback, allowing the subject to take it at his or her own pace, can reliably lead to learning. Up to a point this is absolutely true, especially in the ‘spaced’ form in which Skinner actually presented it rather than the simplistic caricature demonstrated in the article. For some kinds of rote learning, the effects of which are easily observed, small chunks and immediate feedback are a very good idea indeed. There are good reasons for this that cognitivist psychologists and constructivist thinkers had also hit upon long before Skinner came on the scene. Although not the archetypal behaviourist way of doing things, it works particularly well if that feedback is innate to the task rather than extrinsically imposed. For instance, staying upright on a bicycle, being able to recite lines in a play, being able to play a piece of music, building a program that does what it should, or writing a satisfying piece of work, are immediate forms of feedback that are intrinsic to the process. This is the form that deMuth uses in the article: he self-referentially demonstrates the effectiveness of the approach by leaving ever larger chunks out of the key terms it employs. In doing this he is actually relying on a cognitivist model of what motivates us (in this case, achievable challenges) rather than a purely behaviourist model of reward and punishment, so it’s not the greatest example of the effectiveness of behaviourism. He is not exactly an expert in the field. Behaviourists also hit on a few other good tricks, more by luck than design. It is absolutely true that putting people in control of the pace of their own learning works very well, both for obvious common-sense reasons (we don’t all learn the same things at the same speed) and for motivation: a sense of being in control is central to intrinsic motivation. This is not a behaviourist notion, but it happens to be true.

The most problematic outcome of behaviourist thinking, that follows from its wilful ignorance of internal motivations and stimuli, lies in the use of rewards and punishments to drive learning, using extrinsic motivation as though we were all pigeons. The big trouble with the reward/punishment idea is that extrinsic motivation actually eliminates intrinsic motivation, which means that a reward/punishment model is positively harmful to effective learning. There have been countless studies and experiments that show this from Deci, Ryan, Kohn and very many others. I am quite taken by a recent paper on the subject which rather neatly shows its effects using 10,000 West Point cadets tracked over 14 years, that summarizes some of the classic research quite well, as well as adding its own compelling evidence. By nature humans love to learn and enjoy achievable challenges but, if you beat or reward that love out of them for long enough, they will stop wanting to do so. The big lesson that we learn from extrinsic rewards and punishments is that learning is done to gain rewards that have no connection with that learning, or (just as bad) to avoid punishment. We also, in passing, learn that the purveyors of those rewards and punishments have power over us.

If it actually worked then even this ugly power trip might be worth it, though I have strong reservations about the ultimate value of teaching people to bow down to authority figures without question. Unfortunately, many studies have demonstrated unequivocally that, though such methods may result in short-term gains that may be sufficient (if not particularly efficient) to pass the big sticks and carrots of exams designed to test behaviour, learning this way does not persist, especially if no attention is paid to meaning, value and connections between things. This accords with common sense and experience. If we are taught that the value of what we are learning is to pass a test or get a grade then, once we have achieved that, it is perfectly natural to promptly forget it. It’s much like remembering your hotel room number: very important as long as you stay there, completely irrelevant when you leave, and therefore promptly forgotten.

Learning that persists is learning that we can continue to use, that relates to our goals, the things we want or need to do, that relates to our social context, that we can apply and that has meaning and value to us because of who we are, where we come from, what we want to do, the communities that bind us, and who we want to be. For this kind of learning, self-pacing, small chunks of increasing complexity and fast feedback can be extremely useful tools (if far from being the only ones) but the point is that it cannot be done effectively in isolation and especially not under the control of someone else. Values and meaning are not in nor can they be usefully described by the behaviourist vocabulary, but they are exactly what deMuth’s reviled educators, against the odds and against the flow of a system that is designed to work in total opposition to them, are trying to foster. At least, the good ones are doing that. Too many of us are buckled down by the system that thwarts us by standardizing learning, trying to make us teach the same things in the same temporal and physical/virtual space, at the same time, over fixed periods, without any thought for the reasons it might be worthwhile to people or their individual and unique needs. It is no wonder that education has one of the highest dropout rates of any profession. DeMuth rightly attacks education but wrongly attacks educators.

The failure of educational systems

As long as we have abominations like the core curriculum, obligatory courses with defined objectives, or coarse-grained programs that ignore individual needs, that make it a requirement to learn a specified body of facts and skills regardless of their personal value or interest to us, this will ever be so. Unless we can devise ways of doing education that will be meaningful, applicable and valuable to the individuals that are learning, without extrinsic rewards or punishments, we have failed. If we teach students that the purpose of learning is to pass tests (or receive some other extrinsic reward or punishment), we will have doubly failed, because we will have made it harder for them to learn anything ever again and, in all probability, will have fostered an aversion to what might otherwise have been an important and interesting thing if it were learned at the right time. Skinnerian teaching machines deployed without addressing these fundamental problems will simply reinforce the same old patterns, making things worse, far worse, than before.

When used to support personally and socially meaningful goals, some behaviourist methods can have limited value, though none of those methods are unique to behaviourism and most come with important provisos and modifiers. Practice can be very good for acquiring a wide range of skills, especially when interleaved and spaced (those studying for exams or learning to play a musical instrument would do well to take note of this), learning things when, how and at what pace we wish to learn them is crucial, and we do need to take things a little at a time. Some such skills are foundational and, once learned, can become self-sustaining and supportive of intrinsically motivated learning: reading and writing, for example, or arithmetic. But behaviourism is not right just because it made a few hits. DeMuth wants to improve literacy (good) but he seeks to improve it through behaviourist methods (very bad) and measure it by standardized tests (very, very bad). This is a bit like assuming that the purpose of the army is to kill people, and therefore providing all soldiers with nuclear weapons. It is putting the cart before the horse. 

The purpose of education is not to pass tests but, along with sustaining some cultural continuity, to help people both to learn and to continue to learn. Behaviourist methods may achieve short-term testing goals but are singularly poor at fostering long-term learning and are positively antagonistic to lifelong learning. They encourage a dependent and submissive attitude and stamp on critical or creative thought. We should let them rest in peace.