I was not able to attend ResearchEd2013 back in September 2013; but ever since then I have been meaning (and not finding the time) to comment on the outcomes of the conference, which were conscientiously videoed and posted to the web by Leon Cych. The conference was organised by Tom Bennett to highlight the importance of (and problems with) current research in education. This was a few months after he had himself published Teacher Proof, mentioned in my earlier post, Why teachers don’t know best.
It struck me that while the attack on quack theories was sound, the conclusions reached in Teacher Proof about the nature of the expertise of teachers were not well justified. Indeed, they seemed to me to be bizarrely at odds with the advertised prospectus of the Research Ed conference.
The meat in the Teacher Proof sandwich is an exposé of a long list of quack theories of learning, include learning styles, twenty-first century skills, group-work, and Brain Gym. Against these theories, Bennett mounts a useful and carefully argued attack, even if its relentless, metaphor-rich humour might sometimes seems more appropriate to a listless lower sixth than the adult reader.
The first section of the book provides an introduction to epistemology (the philosophy of knowledge) and research. Although many might see this as a diversion, I think that such a review is much needed and is in general well done. Even so, it contains a couple of questionable sections. These are in turn responsible for what strikes me as an unsatisfactory conclusion, in which Bennett asserts that teachers should beware all theories of learning and trust instead to their own experience and craft. It is this conclusion (oddly incompatible with the prospectus behind ResearchEd2013) that I wish to challenge; and I will start by explaining my criticisms of Part I, which contains the overview of the theory of knowledge.
As often as you hear some up-and-coming politician come onto the Today programme and assert confidently that A causes B because the two are correlated, so too do you frequently hear it said by people like Tom Bennett that “correlation does not imply causation”.I am describing, of course, the difference between correlation and causation, a reef of rocks of which every good scientist is wary. Correlation is when events happen to accompany each other, without necessarily being the cause of each other. In my fabulous and shameful example, there is merely a correlation between the day of the week and the pounding in my head.
The problem is that the popular aphorism “correlation does not imply causation” is, strictly speaking, not true. If you think that my position here is so absurd as not to be worth reading any further, have a look at the section titled The Slogans at http://www.psy.gla.ac.uk/~steve/best/correlation.html, which should at least persuade you that the issue is rather more complicated than is commonly acknowledged.
The section in which Tom Bennett discusses this issue follows a review of David Hume’s discussion of correlation and causation in his Enquiry concerning Human Understanding. Hume pointed out that we do not really understand what causation is—we just think we understand what he called the “necessary connexion” between two events.
In fact, all we ever see of causality is the fact that one event (e.g. throwing a brick at a greenhouse) is invariably followed by another (the greenhouse glass breaking). This is what Hume called a “constant conjunction” and what we would call a reliable “correlation”. We can of course create intellectual models that attempt to explain why the glass breaks. But however plausible the models might seem to us, the models do not give any hard evidence for their own validity (climate scientists take note). The only way we can ever gather hard evidence for our theories is through experimentation—in this case, by throwing lots of bricks at lots of glass panes and observing the correlation (or lack of correlation) between the two phenomena.
Those (including Tom Bennett) who say that correlation does not imply causation miss the key significance of Hume’s argument. While the popular aphorism suggests that causation and correlation are completely different types of relationship, the truth is that as far as we can ever know, causation is a sort of correlation. This means that we can confidently say that wherever there is a causal relationship, there too is a correlation.
What is almost never acknowledged is that we can also say the opposite—where there is a correlation, there is likely to be causation—subject to two important caveats. The first (which explains the “correlation does not imply causation” aphorism) is that a correlation between A and B does not imply that A causes B. B might cause A or C might cause both A and B—these three cases are summarised in the diagram below.
So it would be correct to say that “correlation between A and B does not imply that A causes B”. But that is not what the aphorism says. It is not true to say that “correlation does not imply causation” because, even if we are not quite sure what is causing what, we can be pretty confident (depending only on the strength of the correlation) that causation is at work somewhere in the vicinity of the correlation. It is as though Miss Marple has found the bloodied body of the butler in the kitchen: she cannot be sure that he was killed by the cook, but she can be pretty confident that there has been foul play on someone’s part.
The second caveat is to emphasise the word “imply”. A correlation never proves absolutely any causal relationship because it is always possible that if you carry on experimenting, you will find a particular case where the conjunction does not occur. But the more you find that two phenomena always seem to accompany each other, the less likely it is that you will find an exception to the rule. In the end, the likelihood that you will ever find an exception becomes so small that it can safely be ignored for most of the time.
The importance of this point (which may strike many as rather pedantic) is that the popular use of the “correlation does not imply causation” aphorism suggests that research (which necessarily shows only correlation) has nothing to say about causality and is therefore, at best, a superficial exercise. In fact, the very opposite is the case: research into correlation is the only means we have of demonstrating causal relationships.
Nor should anyone be too concerned that a single correlation cannot in itself show what is causing what: this level of precision can be achieved by triangulating multiple correlations. In order to establish whether a third phenomenon, C, is responsible for both A and B, you need only to conduct the experiment in an environment in which C is not present: if A and B are still correlated, then you can then be sure that C is not responsible. Although you have still not proved the negative “there is no other confounding factor, other than A and B, that is responsible for the correlation between A and B”, it is possible to disprove, experimentally, the significance of any plausible third factor that might be proposed. And this should satisfy us because no empirical evidence of any kind ever gives us certainty: only extreme likelihood.
Once you have established that the correlation is not caused by any plausible third phenomenon, it is simple enough to show which of A and B is the cause and which is the consequence. It is an absolute rule that cause always precedes consequence—so that by showing which of A and B occurs first, you can establish whether A is causing B or B is causing A. It is also possible, of course, both that A causes B and B causes A—in which case you have feedback effect.
The assumption that “correlation does not imply causation” may be partly responsible for what I suggest is Tom Bennett’s mistaken conclusion that teachers should not trust the conclusions of educational research, which seeks to demonstrate causality by pointing to correlations—just like all other forms of research.
The main thesis of Tom Bennett’s Teacher Proof is that social science is not the same as real science because it is laden with value judgements. The implication is that value judgements are inevitably subjective and can therefore never be said to be objectively true, as can factual or logical assertions. This is another very common view.My position could broadly be described as Positivism; this was the theory that the only meaningful things that science could and should describe were empirically verifiable; it posits the existence of an external world that is objective and not mind-dependent. In other words, there’s a real world that we can observe and agree upon. Social science gets a beating from positivism because it talks so much about matters of values, perspective and subjective belief. While we can agree on the mass of a brick, we can’t find such easy consensus in describing the fairest way to distribute resources.
I agree with Tom Bennett if this is meant to be an observation on the character of much social science that is practiced. But I do not agree that value is necessarily subjective.
It is worth beginning this discussion by distinguishing between ethical judgements (which concern what we should do) from aesthetic judgements (which concern matters of beauty and taste). Education theory concerns what we should do and the observation that this involves subjective value ties in with the perception of many teachers that their job is inevitably value-laden and subjective.
Both aesthetics and ethics can be seen as exercises in psychology. There is a common thought experiment which comes in various flavours, but which in each cases asks whether it is right to kill a single person in order to save many people. Generally, people become very queasy about the ethics of pushing a particular, very fat person who happens to be standing next to you onto the railway track in front of a runaway train in order to save the lives a seven people tied to the tracks further down, even when the utilitarian argument clearly suggests that this would indeed be the correct course of action. To my mind, the only plausible way to square this circle is to say that it is ingrained into us at a deep, archetypal level, that the pushing of people onto railway tracks is morally wrong, even when our intellects argue otherwise. Such an argument makes very little distinction between ethics and aesthetics: the reason we like a symmetrical face or a scene of a hamlet in a fertile, well-watered valley are similarly ingrained into our collective unconscious for evolutionary reasons. From these perspectives, both ethics and aesthetics become branches of psychology (and maybe of evolutionary biology) and do not involve the direct consideration of either the good or the beautiful.
Such an approach, at least in the case of ethics, becomes increasingly unsatisfactory in an age of technological change, when traditional, instinctive behaviours are often unreliable guides to how one should act in modern society. The most obvious example is how the expression of male aggression, a fundamental instinct which is often constructive in the functioning of animal societies, needs to be modified in modern societies in which we become increasingly dependent on those around us. So the question that lies behind all ethics, “how should we act”, cannot be answered by looking merely at how nature has designed us to act but at how reason dictates that we should act in situations which nature may not have anticpated. And the criteria that we must invariably use in order to answer that question are those proposed to us by the theory of utilitarianism: what are the consequences of our actions?
Consequences may be divided into two sorts: intended and unintended. Intended consequences are also known as “ends”, while the actions that we take to achieve those ends are our “means”. It is not true that “the ends justify the means” but it is true that the consequences justify the cause. The problem is that the unintended consequences are always much more numerous than the intended consequences, mainly because there is no end to them: the consequences of our actions stretch from here to the end of time and almost all of them are unintended.
We keep our sanity in front of this vertiginous drop by accepting a sort of limited liability, ethically speaking. If we rescue someone from drowning, we consider that we have done a good deed, even if it turns out that we have rescued a mass murderer, who is subsequently responsible for countless deaths. Even though each one of these deaths is a direct consequence of our supposedly good deed, we do not take responsibility for it. The moral baton of responsibility has been passed on.
Each leg of this moral relay race could be condidered as a link in a chain; each link has its own means, its own ends, maybe its own short-term unintended consequences, and its own ethical status. And then the focus moves to the next link (or links, as all causal chains are constantly branching). But if you were to follow the central, intended, purposeful chain, the ends of one link become the means of the next.
This point was illustrated by Aristotle’s in the Nichomachean Ethics, in a passage that I first explored in my earlier post, Aritstotle’s saddle-maker. The grid below summarises a chain of ends and means in Aristotle’s example, in which every action or instrument appears as both ends and means.
|Saddle-maker||Leather and tools||Saddle|
|General||Cavalry charge||Winning a battle|
|Politician||Winning a battle||Supremacy of the city|
At every stage of this chain of interdependent purposes, the success of the means being employed can be objectively tested. It can be shown whether the better saddle is made with metal or bronze tools, with pig hide or cow hide, by Alexander or by Alexis. “Ah!”, you say, “but you can’t show, objectively speaking, whether making saddles is a good thing to do with your time—maybe our saddle-maker would do better if he took to growing lettuces instead”. Well, you can show, objectively speaking, that making saddles is a good way to spend your time if you want to have the capability to execute effective cavalry charges in battle. And that too, can be shown to be valuable for other reasons. At every stage, the truth of an ethical proposition (i.e. a proposition about how we should behave) is demonstrable subject to particular conditions being met. The ethics of every link in the chain are conditional, which is not the same as being subjective.
You are not yet impressed by this argument, which you will point out must always hinge on the acceptance of some ultimate objective. In this case, it is the “supremacy of the city”, which someone who believes in peaceful co-existence may very well reject. I will admit that there is some force to this point of view—but the force is much less than you might assume, for four reasons.
- In a complex society, the responsibility for deciding ultimate ends often rests with someone else (such as the customer). If the saddle-maker thinks that making saddles is not a worthy occupation, he should get himself another job. So long as he stays a saddle-maker, the ethics of his situation require him to make good saddles.
- David Hume (again) suggested that although it is true that we could not objectively prove the legitimacy of our ultimate ends, in practice this is hardly ever necessary as almost everyone’s ultimate ends are to be happy. It follows that almost all ethical disagreements concern means, the effectiveness of which can be objectively demonstrated and disproved.
- A very similar objection can be made to all cases of deduction, by which a conclusion may be drawn from two premises. The conclusion is only true if the premises are true and the truth of the premises can only be shown by further deductions, ad infinitum. Yet this has not persuaded anyone that the validity of the process of deduction is subjective.
- In practice, most people who assert that ethics are subjective are not prepared to put this theory into practice when they are asked to accept, for example, that the ethical status of genocide and torture is just a matter of personal viewpoint.
So although it is important to recognise that all judgements (ethical, factual or rational) involve assumptions, I do not agree that ethical judgements involve a greater degree of subjectivity than other forms of judgement. If we can agree that we want children to learn certain sorts of things, then it is comfortably within the capacity of quantitive educational research, given certain sorts of children with certain sorts of prior attainment, to show which teaching methods are more effective and which teaching methods are less effective at inducing bringing about that learning.
The third assertion in Tom Bennett’s introduction to epistemology that is in my view misleading is that “Experience trumps theory every time”. He states this rule in the following passage.Here’s what I believe; this informs everything I have learned in teaching after a decade: Experience trumps theory every time. There are rules of the classroom: broadly productive mechanisms that are useful in directing and educating a room. But as I have said, the diversity of human experience ensures that few specific, micro-management techniques persist in being universally efficacious. An example would be: children are deterred from poor behaviour by sanctions, and generally are encouraged by rewards. But because people (and children) are complex, multi-layered, complex beasts of illogic, emotion, background, reason and possibly fre will, they are not predictable I anything more than broad brushstrokes.
As a general rule, “experience trumps theory every time” is unsound because without theory, we have no way of interpreting our experience. Even before our sense data is presented to our conscious minds, it has already been interpreted according to a whole range of automatic, sub-conscious processes, all of which encapsulate some kind of theory. At a conscious level too, we can only make sense of our experience because of how it fits (or sometimes challenges) our particular world view. Without the filter of theory to allow us to make sense of the world around us, experience is just noise.
It may be that Bennett does not mean this as a general, epistemological theory, but rather as a commentary on teaching as a particular case. And there are two particular characteristics of teaching that may need to be considered:
- the failure of education researchers so far to develop satisfactory theories of teaching and learning;
- the complexity and variability of humans.
While I agree with Tom Bennett on the first point, it does not amount to an argument against theory in principle or against the need develop better theories of learning than we may have at present. It would be a mistake to dismiss medicine as a methodology, just because some people are quacks or because we do not at the moment have a cure for every disease.
In the passage quoted above, it is on the second point that Tom Bennett pins his colours. He criticises theory on account of the variability of human nature and experience. But his implication that theory attempts to impose one-size-fits-all solutions to complex problems has no validity at all. The science involved in civil engineering does not propose that all bridges should be identical—it gives a range of rules that help decide how a bridge may be constructed depending on the nature of the site, the volume of traffic and the client’s preferred aesthetic. Theory is abstract: the application of any theory needs to be context-dependent and it is an extremely poor theory that does not take into account the fact that context varies.
In spite of the useful critique of quack learning theories, Teacher Proof fumbles the main argument in the first and third parts, ending up with a superficial conclusion which does not help us find useful solutions to our problems.
By telling teachers to rely on their own experience, Bennett does very much more than criticise current research: he dismisses all research and all theories. In this respect, the conclusion is not only wrong but also absurd. Because individual teachers who follow Bennett’s formula of trusting to their own experience are not doing without theory: they too have theories too that guide their actions.
The difference is not between theory and experience but between on the one hand, explicitly articulated, openly contested, public theory; and on the other by poorly articulated, unchallenged, private theory. Don’t even ring the bell for the beginning of round one: there is no contest between the two as a basis for professional practice. If private intuition were so reliable, teachers would not have so easily fallen prey to quack theories in the first place. The answer to bad science is good science; not a return to a romanticised vision of the master craftsman operating on the basis of a mystical, inexplicable capacity to effect change.
I would make a similar criticism of his treatment of education technology in a chapter entitled “Techno, techno, techno, TECHNO: digital natives in flipped classrooms”. While I agree with Tom Bennett’s criticisms of much of the current implementation education technology, this is an argument against technology as currently implemented, not an argument against technology in principle. At the beginning of his chapter entitled “What can teachers do?”, Bennett says:If there’s one thing I’ve found from my job, it is that teaching is a lot of hard work, but it isn’t expensive, in the main. It doesn’t require a box of Buck Rogers tech, it really doesn’t.
Although it is clear that no-one requires “a box of Buck Rogers tech” or indeed more “dreadful one-day conferences”, it is an underhand argument to suggest that because no-one wants pointless tech and bad training, they may not need useful tech and quality training. And if Bennett thinks that teaching based on a model of teacher-as-independent craftsman is not expensive, then he needs to look in the mirror from time to time.
It may well be that Tom Bennett would already be prepared to accept at least some of the points that I have made in this post. In his introduction to the ResearchEd2013 conference, he admits that:In fact I repeated time and time again something I wrote. I said “experience trumps evidence every time”. I don’t quite agree with that any more.
But this certainly does not amount to a recantation because in the same speech, Bennett says:You know what? Despite everything I have said so far, I don’t think everything, I don’t think even half, of matters within schools and education can be settled by research. I think a lot of it is intuitive. A lot of it is craft…I am a better teacher now than I was ten years ago because I understand the subtleties of the invisible interactions and sometimes, when you don’t say something. And I think that is very hard to codify. It’s very hard to metrify.
My disagreement with this passage lies not in the fact that teaching is difficult and its subtleties may take many years to master. My disagreement lies in the assertion that those subtleties cannot be expressed by science and their effects measured and demonstrated objectively. But it is to this question of measurement—or metrification, as Bennett puts it—that I will turn in future posts.
As he says in the introduction to Teacher Proof, Tom Bennett wrote the book largely because he was so appalled at the nonsense that he was fed during his teacher training. As a way of venting some of that anger and identifying the problems that have beset educational research, the book works well. But in respect of arguing for a solution to those problems, the book falls short and, in my view, is in danger of pointing us in precisely the wrong direction.
As discussed in previous posts on this blog, research has shown repeatedly that good teaching is by a country mile the most important contributor to better learning. However, it is in the next step of the argument (a step so small that people do not even notice that they are taking it) when people trip up. They assume that better teaching can only be achieved by recruiting and training better teachers.
Bennett’s conclusion—that the skill of teaching is necessarily a private matter, based on private knowledge, intuition and experience—reinforces this unjustified assumption, which is characteristic of a pre-industrial, pre-scientific, craft economy.
What the ancient Greeks understood is the importance of what they called “giving a rational account” of the world around them, and in particular, giving a rational account of their expertise. Socrates asks the orator Gorgias, “In what does your expertise consist?”—and when Gorgias fails to give him a defensible answer, he concludes that Gorgias’ claim to expertise is a sham.
Only when we can answer Socrates’ question and only when our answer is detailed and justified by evidence can new recruits to the teaching profession be properly trained. Only then can those aspects of teaching that are not dependent on human teachers be delivered efficiently, consistently and at scale by appropriate course materials and programmes of study. Only then can schools be reliably inspected. Only then can we make effective use of the time of highly skilled and expensive graduate teachers.
To answer Socrates’ question by saying that your expertise is a matter of private intuition, experience and craft is to give no answer at all.