Let’s not trust the experts for a change

There’s a flurry of surprise in the media at the suggestion that “neutrinos travel faster than light”. It seems many of us are surprised that scientists are surprised. Again and again, the media make the mistake of thinking that science is a very trustworthy body of knowledge, instead of a very risky, hypothetical, often internally inconsistent body of guesswork.

Why? – I think it has something to do with the social status we accord scientists, and experts in general. And the urge to accord an elite special social status seems like a nearly universal human failing. Most human societies seem to have notable figures – such as “holy men” or experts in subjects that are arcane to the average person – who are not simply consulted for their opinions, but are entrusted with decision-making powers on the basis of their expertise. These experts are not entirely to blame for their own status: ordinary people are generally uncomfortable with uncertainty, and hate making decisions in a vacuum. So they tend to abdicate the responsibility – and avoid the anxiety – of thinking for themselves, by passing decision-making powers to others. For their part, most experts are happy to accept this honour. The “activists” among them actively seize the honour. For that reason among others, academics naturally gravitate towards the role of “expert” or “guardian of opinions” in society.

That may be good for academics, but I think it’s bad for individual development, bad for truth, and bad for knowledge.

And it’s bad epistemology. An expert is someone who is unusually familiar with his subject (i.e. a set of theories) rather than someone who makes unusually reliable judgements about the subject matter (i.e. the reality the theories purport to describe). To illustrate this, take economists. Some of them passionately believe that Hayek was right, while others passionately believe that Keynes was right. Both no doubt have endless arcane details of their respective theories at their fingertips. Whichever side happens to be right, the self-assurance of economists is evidently no measure of the reliability of their opinions. The same applies to experts in theology. However familiar they may be with the “five ways” of Aquinas or the “seven pillars” of Islam, they are the last people anyone should consult on the question of God’s existence.

Or take philosophers. A utilitarian like Peter Singer can make a decent case for euthanasia, and a non-utilitarian like Roger Scruton can make a decent case against it. Evidently, an “expert in moral issues” is someone who spends a lot of time thinking about moral issues, not someone whose moral opinions are more reliable than the moral opinions of others. (If they were more reliable, there would be more convergence.)

Experts such as scientists do not have more reliable opinions, but more penetrating opinions. In other words, they theorize about the hidden structure of reality, but the more hidden is the reality they theorize about, the riskier their theories get. The history of science is a series of ever more penetrating theories, almost all of which are eventually rejected as false. But lack of certainty doesn’t matter here – the beauty of science lies in its penetrating power rather than its certainty.

The one place where certainly (or at least reliability) really does matter is in action. We act to achieve goals, and we should be guided by trustworthy beliefs about how to achieve them. To strive for the required level of trustworthiness, ordinary non-expert members of the public should listen carefully to expert opinion, on all sides, but make judgements for themselves. As Edmund Burke noted, trustworthy judgements are more likely to be made in the everyday realm of common sense than in the rarefied air of arcane expert opinion. That is why juries of ordinary people make decisions in law courts rather than expert witnesses. Jury members listen to expert opinions, of course, before they decide. One of the most important functions of the jury is to force expert witnesses and legal representatives to make their case in layman’s terms, in other words to drag the discussion down to Earth, where reliable judgements can be made.

Let’s discuss scientific method, not politics!

Earlier today, I tweeted how I “wish climate sceptics would keep politics out of debate! Problem is methodology: climate science is inductivist, not hypothetico-deductive”.

I guess that must have sounded very pretentious and unclear, so I’ll now try to explain in simpler terms what I’m talking about.

A standard example of induction is the jump from “all of the swans I’ve see so far have been white” to “all swans are white”. Induction is the form of reasoning in which we jump from noting a property (in this example, white) of a limited number of individuals (swans I’ve seen so far) to attributing the same property to all of the members of the same class (all swans).

Induction is essentially extrapolation. So just a moment ago, I extrapolated the property of being white from the sub-class of the swans I’ve seen so far to the larger class of all swans.

Most of us are familiar with the use of extrapolation in statistics, where the relative frequency of some characteristic in a sample is assumed to match that of the entire class from which the sample is taken. For example, we might jump from “roughly half of Irish people are male” to “roughly half of all humans are male”. Or from “10% of Irish people have red hair” to “10% of all humans have red hair”. Or from “25% of those who answered the survey will vote for the X Party” to “25% of the entire electorate will vote for the X Party”.

Induction is a very useful type of guessing, and it can be very reliable. It’s reliable when the smaller sub-class from which the jump is made is genuinely representative of the larger class. So although the example above of white swans is not reliable (because there are some black swans), the superficially similar jump from “all of the emeralds I’ve seen so far have been green” to “all emeralds are green” is indeed reliable. The greenness of emeralds is not an accident like the whiteness of swans, but an essential feature of emeralds: all emeralds contain chromium, because otherwise they wouldn’t count as emeralds – and the presence of chromium makes them green.

Indispensable though induction is, and although we can no more give it up than we can give up breathing (as Hume remarked) it is not the characteristic type of guessing found in science.

In science, guesses — otherwise known as hypotheses — are the logical starting point rather than the logical result of careful types of observation. For example, Newton guessed that all massive objects attract all other massive objects, with a force that depends on their mass and how far they are apart. No one had thought of that before, so the many observations that many people had already made did not imply or even suggest that such a force existed. If they had, someone else would have come up with the idea long before Newton. But no one did.

After coming up with his guess, Newton then went to a lot of trouble to work out the consequences of his guess, using mathematics that is still rightly regarded as one of the greatest achievements of any human mind. He was able to work out how the Moon, planets and other falling bodies would move if they were subject to the force he had guessed about. And sure enough, careful observations revealed that they did indeed move as his guess predicted they would move. The consequences of his guess were confirmed by observation.

When the observational consequences of a hypothesis are confirmed like that, the hypothesis itself is not “proved”, nor even “confirmed” as the predicted observations were. That is because the observations do not imply the hypothesis. It is exactly the other way around: the hypothesis implies observations. The predicted observations follow by deduction (i.e. the sort of strict, non-guessing logic seen in mathematics) from the hypothesis in concert with some other assumptions. If the observations are confirmed, in other words if the prediction turns out to be true, the hypothesis is at best “corroborated”. That is, we have a reason to think the guess was “on to something”. If the predicted observations were really unexpected, the hypothesis “made it over a hurdle”, and the higher the hurdle, the more the hypothesis comes up smelling of roses. Although, I repeat, it is never conclusively proved.

Because the observational consequences are deduced from the hypothesis by deduction, this pattern of testing is called the “hypothetico-deductive method”. But no one should think scientific hypotheses are deduced from anything, least of all “data”. Scientific hypotheses are tested, not deduced. The real test occurs when actual observations are made, and found to agree (good) or disagree (bad) with what the hypothesis said should be observed.

In the seventeenth century, scientists and philosophers such as Galileo and Boyle began to see that this use of hypotheses was the characteristic form of reasoning in science. But some men of letters such as Francis Bacon didn’t see it, and thought instead that science was nothing more than the rigorous application of induction. This remains a very influential idea, especially among statisticians, sociologists, psychologists, in medicine, and among those who belong to the vague school of thought called “positivism”. But it’s wrong.

I would argue that that’s essentially what’s wrong with climate science too. In effect, they’re hoping computers models will embody a hypothesis about how the Earth’s climate will unfold over the course of time, derived by extrapolation from how the climate has already unfolded over time. This is a forlorn hope, borne out of a misguided understanding of how science works.

Discussion of climate science has been perverted by politically-minded people, on both sides, who claim the theory or the scepticism is just a conspiracy — to crank capitalism into a higher gear, to raise taxes, or whatever. Well, I’ve had enough of all that. Let’s talk about scientific method instead.

It is remarkable how many professional commentators on science happily condemn their opponents as “nutters” or political stooges while apparently giving no thought to the methodology of science, the very thing that makes for science or pseudo-science. I’m talking to Geoffrey Lean of the Daily Telegraph and Ben Goldacre of the Guardian, if you’re listening!