The myth of positivism


If you spend any time discussing evidence and education you will inevitably, at some point, be called a ‘positivist’. You may even stand accused of ‘scientism’. It is worth examining these claims in some detail because they are generally false and yet they represent quite a successful strategy for those who make them. The edifice of nouny waffle that constitutes so much education research needs an immune system to survive and these terms represent its killer T cells.

Positivism is a philosophical tradition that asserts the supremacy of observation, experiment, logic and reason over other forms of knowing. It therefore rejects metaphysics and religion and tends to take a utilitarian approach to morality. It seems unlikely that many of those who make claims about evidence in education would accept all of these tenets. Instead, it seems that a form of faux positivism has been developed, specifically for mounting attacks on people who try to use evidence in the social sciences. It goes something like this:

Positivists believe that there are clear and specific answers to questions and that these can be established using the scientific method – that the world is black and white, deterministic and can be fully and concretely described. While this is true for the physical sciences, it is a hopelessly naive way to deal with social sciences because social sciences involve human behaviour. Social scientists understand that different interpretations and perspectives mean that there are no simple, easy and generalisable answers.

Not only does this perspective misunderstand the role of the scientific method in the social sciences, it completely misrepresents the physical sciences. To understand why, it is worth briefly examining the second term in the armoury, scientism. This generally means an appeal to science as a source of authority when such an appeal is not warranted. Here is an example:

  1. A high sugar intake causes health problems
  2. A sugar tax will reduce average sugar intake
  3. We should introduce a sugar tax

The first two statements in the argument are scientific in nature in that they are potentially testable. They make claims about the world that could be confirmed or refuted by examining evidence. Collecting this evidence may not be easy. In this instance, you probably could not run controlled experiments and so you would have to look for correlations and these would be contestable. Nevertheless, there is the potential to examine such claims in a methodical way.

The third statement is not scientific in nature. It is about what we should do. This is fundamentally a moral and political question. For instance, there may be those who object to trying to manipulate behaviour in this way, even if they believe it would be effective. Anyone who attempted to claim that the third statement was scientific and that science demonstrates that we should introduce a sugar tax would be guilty of scientism. They have overreached the bounds of science. In this sense, scientism is close in meaning to the proper sense of positivism. And yet those who profess to object to supposed positivism in education research might not even notice the problem in this type of argument.

They may also not notice another issue. When a scientist claims that the evidence shows a sugar tax will reduce average sugar intake, they are making general claims about the population as a whole. They are not making specific claims about exactly how this tax would affect every individual or that the effect would be exactly the same on every individual. It is not necessary to do so. You can still make general claims without specifying what would happen in every individual case.

Similarly, when a reading researcher claims that a particular method of teaching reading is, on average, more effective than the alternatives, it is not necessary for the researcher to be able to specify that this will be true for every single individual or to be able to predict exactly how this method will impact on every single individual. Pointing out that the effects on specific individuals are unpredictable does not refute the original claim.

And it is not just social science that is like this. It is the rule rather than the exception across the sciences. The idea that the physical sciences are completely uniform, predictable, concrete and complete descriptions of nature is just plain wrong.

Science works by creating a model, using that model to make predictions and then performing experiments or making observations to see whether those predictions are accurate or not. A model that is pretty good at generating accurate predictions becomes a theory. Models contain abstractions and are necessarily incomplete descriptions of the world. They often posit the existence of something that cannot be directly observed. Probably the most famous example is Darwin’s theory of evolution which could not explain exactly how genetic information was passed from a parent to a child.

Consider then a model used by cognitive scientists; that the mind contains a limited working memory and an effectively limitless long-term memory. This is the central model of cognitive load theory. It does not matter whether this is a complete description of how the brain works. It does not matter that it is impossible for long-term memory to be literally limitless. It does not matter that we cannot point to the bits of the brain that house these various components. It does not matter that individuals vary in their working memory capacity. None of these things makes this model fail as a scientific model. Crucially, it can be used to make testable predictions and that is really all that is needed.

But isn’t physics different? When it comes down to actual physical matter, can’t we be certain, at least in principle, as to exactly what is going on? Not really, no.

Take the example of the weather. This is a physical system governed by physical laws that are well known. Yet it is notoriously difficult to predict the weather because small changes in starting conditions lead to massive changes later down the line. It is a ‘chaotic’ system. As a result, we are nowhere near an accurate and complete model of the weather, although we have become better at making imperfect models and using these to make reasonable, short-term predictions.

And physics is fundamentally unpredictable at the smallest level. If I fire individual photons of light through a pair of slits and onto a screen, it is impossible, in principal, to figure out where each photon will land on the screen. However, we can accurately predict the pattern that the sum of these photons will make. This is analogous to demonstrating that, on average, one educational approach is more effective than another while not being able to completely predict the effects it will have on each individual student.

Yes, there are still fundamental differences between both kinds of experiment. Social factors that may impact on the results of an educational trial can be far harder to control than the kinds of factors that may impact on a physics experiment involving light, but that does not amount to the need for an entirely different model of what it means to know something or of how to demonstrate whether a proposition is true. It does not mean we can simply reject the all of the accumulated evidence about how people learn by labeling is as ‘positivistic’. Robust, repeatable findings cannot simply be magicked away.

11 thoughts on “The myth of positivism

  1. “Similarly, when a reading researcher claims that a particular method of teaching reading is, on average, more effective than the alternatives, it is not necessary for the researcher to be able to specify that this will be true for every single individual or to be able to predict exactly how this method will impact on every single individual. Pointing out that the effects on specific individuals are unpredictable does not refute the original claim.”

    Any statistician worthy of the name will not only state the change in the average, but also the variance. A particular method of teaching which raises the average but also increases the variance in reading ability might be one which benefits the good readers a lot, but might be quite lacklustre or even have a negative effect on the mediocre readers. So in a sense researchers are already taking into account the effects on specific subpopulations.

  2. As an engineer in an education doctoral program, I made my epistomology professors think twice about their descriptions of positivism. As a result, we have great discussions that really delved into the nature of knowledge in tradiational physics science.

    The criticism about positivism that I heard in my doctoral program is the lack of acknowledgement that the direction of our scientific research does indeed have biases. There was nothing about not using evidence in social sience research
    For example, in medicine, the male anatomy is the standard and female is the exception. I truly believe that if men gave birth, we would have tests to know exactly when labor would occur. However, unti recently, a woman’s time was not considered a resource.
    Another example: Research in pharmeceuticals is driven by the profit making ability of the drug.
    And then there is the whole eugenic phase of genetic research. Yes, biases sometimes do get corrected over time, but they do exist and need to be acknowledged.

    Also a large difference between social science research and physics research is the treatment of context One of the fundamental tenets of physics is that the laws of physics are true anywhere in the universe. I am not sure if there are any laws of social science that are assumed to work in every culture/context. One of the biggest drivers in education is poverty. However places like Finland have features that minimize this effect so even families at the bottom of their socioeconomic base have good educational effects.
    So your analogy to quantum effects does not really work because every time photons are sent through a double slit, you will get the same diffraction pattern. Everytime an intervention is made in a school, you will not get the same result. Very little high quality ed research will state that a teaching method for reading is the best without an exhaustive description of the context of the students studies and the measurement instrument used. What is measured is and how it is measured is a larger factor in social science research than in physical science.

    Being knowledgable in both physics science and the nature of social science research has shown me how different the two worlds are. One is not better than the other. They study different phenomenon.

    1. The limitations of human working memory are not dependent on context. The resulting finding that explicit teaching methods are superior for teaching academics concepts such as mathematics to relative notices is not dependent on context. And that’s the same for reading. Choosing what to research is affected by bias, that is true. I would not base too many conclusions on the model of Finland because it’s performance on PISA, the measure that brought it attention in the first place, has significantly declined over time.

      1. Thanks for the reply
        I agree on the working memory part – that is physical science, not social science!
        Explicit teaching methods do have a context associated with them because what counts as success is open to discussion along with what counts as explicit. If you send me some papers you are referring to we can have a more specific discussion.
        Finland – it is a complicated situation, but my point remains that there are no sociological “laws” that are assumed to be context free.

  3. Thanks for the reference. Because it is more of review of lit., I will have to dig into the references to get to the contexts. From what I have read, this just revisits the old direct instruction vs discovery argument with discovery meaning completely without any guidance and direct instruction meaning anything else.
    These are very wide ranging categories so yes my context idea does not apply. They are not explicit teaching methods but I will dig in and see about the individual papers. For example I just finished Klahr and Nigam and their description of direct instruction was only the following;
    “For each experiment, the instructor asked the children
    whether or not they thought the design would allow them to ‘‘tell for
    sure’’ whether a variable had an effect on the outcome. Then the
    instructor explained why each of the unconfounded experiments
    uniquely identified the factor that affected the outcome, and why each
    confounded experiment did not.”

    I do not consider that a sufficient desctiption of a teaching method. For example, when the kids answered, what did the instructor say?

    I’ll keep reading. Thanks for the paper…it is a welcomed distration from politics.

    1. This paper is not about comparing completely unguided approaches with explicit teaching, despite the title. The cognitive architecture argument relates to more versus less guided approaches. It may seem like an old argument to you but it is of critical importance in schools, in my experience.

      The Klahr and Nigam experiment is mainly interesting because it demonstrates that even those students who did learn via discovery learning understood the material no better than those who were explicitly taught, casting doubt on one of the benefits claimed for discovery.

  4. A couple of thoughts Greg:

    First –

    How is the sugar tax example distinct from this one:

    1. Students with superior marks in assessment type A statistically enjoy greater life chances
    2. Teaching method X secures students higher marks on average than teaching method Y in assessment type A
    3. We should introduce teaching method X and abandon teaching method Y

    As you put it – ‘Anyone who attempted to claim that the third statement was scientific and that science demonstrates that we should introduce a sugar tax would be guilty of scientism. They have overreached the bounds of science.’
    Is my ‘teaching method’ example also one of scientism?

    Second –

    My understanding of the term ‘positivism’ when used as a pejorative is that it draws from the point above about the natural sciences: because scientific theory is founded on provisional and theoretical abstraction, it would be a mistake to confuse the objects of an empirical theory with reality itself. This is a critique which is just as applicable to ‘hard’ science as to any other kind.

    The anti-positivists are thus making a stronger claim than you discussed in your piece- not only are they saying to empiricists ‘get your tanks off our lawn”, but they are also parking their own tanks on the territory of hard science. This is because they think they have other means of getting to reality which they consider more fundamental, or at least enjoying a different kind of legitimacy. Positivism in this sense is the error of mistaking the objects of scientific theory for reality, and so failing to grasp the importance of non-empirical means of research. To argue with such people, it’s no good saying ‘well, actually, social science is closer to natural science than you think’, since natural science itself is considered to suffer critical limitations as an ultimate means of inquiry.

    Even without a well established model, I can confidently predict that you won’t feel much sympathy for this! I’m not sure I do very much myself either, but I think the principle of charity demands we consider the full picture.

    Joe P

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.