The limits of science
A playground for skeptics
From time-to-time, I stand accused of ‘scientism’. This accusation comes from those who argue that the kinds of quantitative studies I draw upon as evidence for and against competing approaches are inappropriate to the field of education. This is a silly argument, but there we are.
Nevertheless, I am convinced we are in the midst of a pandemic of scientism.
It was Margaret Thatcher, speaking on breakfast television in 1989, who exclaimed, “Advisers advise, ministers decide!” And yet in the intervening years, we have seen the steady rise of advisers appearing to decide. This is what it means when a government minister claims to be following the science when responding to the COVID-19 outbreak. They are outsourcing the responsibility that is rightly theirs to bear, to a civil servant.
Why is this inappropriate? Well, science cannot tell you what you should do. For instance, we may be able to determine, through scientific modeling, that reducing the speed limit to 20 kilometers per hour would save hundreds of lives per year. Does that mean we should do it? Not necessarily. We have to weigh the cost in lives against the cost to travel times, business and general prosperity. How can you weigh lives against economic costs? I don’t know, but it’s clear there is no scientific equation for doing this.
Throughout the global pandemic, science has produced models that have informed decision making. Many of these models were inaccurate in the early stages of the pandemic. This was not the fault of the scientists but the result of a lack of information about a novel virus. We didn’t even know if COVID-19 was mainly spread by contact with surfaces or through the air, and our early guesses about this now appear to be wrong. However, even if all scientific predictions had been largely accurate, science still could not weigh excess deaths due to COVID-19 against the economic harm of largely halting the economy through lockdowns*. Making such a decision is a moral and political decision that is nothing to do with science. You cannot follow the science to make such a choice.
And yet, repeatedly, this is what we have been told.
What normal person wants to decide whether to continue to commit troops to Afghanistan, risking more domestic casualties, or withdraw and enable the country to be taken over by an oppressive force that will persecute women, minorities and opponents? Just as undertakers and slaughterhouse workers complete the tasks that most of would prefer to not think about, politicians have to make cruel decisions, knowing that whatever they decide, they will be morally compromised. Normal people don’t want to make these calls. This is why we have politicians and what distinguishes them from activists. We give them the trappings of power and they make these hard choices for us in return. Passing the buck to ‘the science’ gets them out of their end of the deal.
To make the best use of science, we need to know where it ends. One terminus is the point of weighing competing, qualitatively different priorities. Other boundaries are more fuzzy.
I guess most people think of science as a technical description of the universe - the blueprint and rules for assembly. This is a misconception because it mostly is not.
A famous example in the history of science is the case of Newton’s laws. These laws work well for describing how objects move. Whether such laws satisfy the definition of an explanation is a separate matter and perhaps ultimately comes down to taste. Nevertheless, the laws work well.
It was Einstein who proposed that these laws break down at high speeds (before ultimately developing his ideas into a theory of gravity). But it’s important to understand what we mean by high speeds - things travelling at an appreciable proportion of the speed of light. Light is fast. A plane flying at the speed of light could do eight laps of the Earth per second. Not many things travel at this kind of speed - at least not many things we encounter every day - and so Newton’s laws work well, most of the time.
Historically, physics went from Newton’s laws to a more fundamental theory of which Newton’s laws are a special case. However, a lot of science that matters in our everyday lives does the reverse of this. Often, we know how the constituent parts behave and what we are interested in is deriving approximate ways of predicting the behaviour of complex systems made of these component parts. We are trying to obtain the equivalent of a set of Newton’s laws that, while incomplete, approximate and destined to break down at the extremes, are really good at making predictions about a set of situations we are interested in.
For instance, it is quite possible to know how a virus operates at a cellular level - how it infects cells, which types of cells it is more likely to infect and so on, without such knowledge being particularly helpful in modeling how that virus will spread through a community. Instead, for the second task, we need to generate models based upon approximations and heuristics and test them against reality.
Such models are not a form of fundamental science, but they are science because they are subject to the scientific method of formulating a hypothesis, testing it against reality and then refining the hypothesis into a theory. These models exist in a messy and unsatisfactory in-between world. They are not intended to be full descriptions of reality but to be useful ways of answering a certain class of questions.
The same is probably true for cognitive science. If and when we do finally figure out exactly how neurons form memories, that may be about as useful to the scientist constructing a learning theory as cell biology is to an epidemiologist. We will still need messy theories operating at a different scale that approximate the learning process.
A scientific model is a model, first and foremost, and only an accurate description of the world up to a point. Its usefulness is relative to the questions you want to answer with it. A model is not flawed just because it fails to include something irrelevant to those questions.
To illustrate this point, think of a physical model. Suppose we want to understand how air will flow over a new design of car. We could make a physical model of the car and place it in a wind tunnel. This may not be a perfect model, but it may be good enough to answer the questions we have about air flow.
It would not be particularly clever or sophisticated to point out that the model fails as a complete and accurate representation of the car because, for instance, the doors don’t open and it has no engine. Unless these features are relevant to the wind-tunnel test, they do not need to be replicated in the model.
This is how I feel when people suggest that the model of the mind used by cognitive load theory is flawed because it does not include, say, sensory buffers. Explain to me why it needs to and I may be interested. Simply pointing out that sensory buffers exist is like pointing out that cars have doors.
The problem with scientism is that it is obviously wrong to anyone sitting in an armchair who is motivated to think about the issues for a while. By leaning inappropriately on the authority of science, we diminish that authority. Due to the widespread misconception that science is meant to be a complete, technical description of the world, rather than a set of models, armchair skeptics can then use the ‘no doors’ argument to suggest a million ways in which whatever scientific model they are skeptical about is incomplete. And that’s how you get to denialism.
Which is bad for science and bad for society.
*For what it’s worth, I am not opposed to lockdowns and I think they were a necessary part of combating COVID-19.