Tuesday, April 23, 2019


What do Scientists Really Know? Why the Philosophy Really Matters.

This column does not often stray into the philosophy of science, but there are occasions when it is useful to reflect on just why we consider certain things to be true and with what evidence. Nowhere is this more true than in evaluation of what science tells us on climate change. Given the superficial and inaccurate material that continues to flow from the many political opponents of action on climate issues, it is worth dissecting some of the fallacies and misunderstandings that find their way into popular debate. One of these is the notion that the nature of scientific knowledge is always provisional, with the additional assertion that it cannot therefore be relied upon in matters of policy. This is a misreading of the nature of scientific method and the philosophy of science. It is often combined with the false claim that climate science has lacked predictive power and in particular failed to anticipate the actual warming that we are now observing. These arguments owe little to logic or fact.


A lot of the popular discussion on climate (and other) science stems from low level interpretations of the work of Karl Popper, and in particular the criterion of falsifiability as a necessary condition for any proposition to be considered as science. So it is worth describing briefly the nature and status of Popper's philosophy. Loosely summarised, Popper's essential thesis was that no proposition could ever be finally proved, but that proper science could be defined by our ability to set tests, often of a predictive nature, for it to pass. Hence falsifiability, or the capability of being disproved, was the test of proper science. Popper has always been popular among political commentators, especially on the Right, because he usefully demolished the pretensions to science status of many Marxists. He observed that they failed, or refused, to produce testable, and hence falsifiable, propositions based on their ideas.

These ideas, taken in isolation, characterise science as proceeding by a series of hypotheses, each of which may or will in turn be falsified and replaced by a better  hypothesis. This is the feature that excites the interest of climate science sceptics. The idea (incorrect, as we shall argue later) that scientific propositions are always provisional supposedly implies that they can and will be found to be false, and are therefore not reliable as a basis for policy. Needless to say, this is an attractive argument for anyone faced with what they may regard as an “inconvenient scientific truth”.

So far, so good. Popper made an important contribution, at the very least as a codification of what had always been present in de facto practice of science since the days of Isaac Newton[1]. But his insight is not the whole story, although it is sometimes treated as such in a political context. Nor is it a complete description of how science works, and Popper is certainly not the ultimate court of appeal or the last word on scientific method.

One excellent source on this subject is the work[2] of Brian Davies, referenced below, which gives a much more coherent and comprehensive account of the nature of proof and knowledge in mathematics and science, fully acknowledging the contribution of Popper among others, but putting it in its proper historical, scientific and philosophical perspective. Another interesting article by Richard Lawson[3] makes a further interesting contribution on the application of Popper’s approach to validation of climate science

Popper and Evolution. Scientific understanding of evolution has had an enormous impact on human thought and on practical science since these ideas developed more than 150 years ago. Comparison with climate science, which threatens to be an issue of similar scale, is therefore instructive, not least because there is some overlap[4] between the sceptics on both issues. However, Popper’s judgement on matters scientific is called into question by his earlier failure to recognise the theory of evolution as science under his criterion. JBS Haldane is famously attributed the answer: “Show me a pre-Cambrian rabbit, and my confidence in the theory of evolution is lost.” Popper later and sensibly retracted.

Not all science provides a basis for future prediction. Falsifiability does not depend on and cannot be equated with the ability to make accurate future predictions. Evolutionary biologists may understand perfectly the mechanisms of genetic mutation but there is no way they can predict, or we test, exactly how random mutations might lead over many millennia to the evolution of new species under the same or changed environmental conditions. Similarly, weather, the day-to-day manifestation of climate, is intrinsically unpredictable for more than a few days ahead. But this is much less likely to be true of climate, the composite or “average” of weather over a long period.

Prediction and falsifiability. The reality of science is in practice much more nuanced and complex than a simple move from predictive failure[5] to falsification. Where a prediction made from an apparently well-established theory fails, the first response should normally be to look for what has gone wrong with the data or the experiment, or what other factors have affected the result. It should not be immediate rejection of a well-established theory.  Otherwise the laws of chemistry and physics would be disproved on a daily basis in school laboratories.

At a higher level, discrepancies in the orbit of Uranus revealed the existence of Neptune, not the failure of Newton’s Laws. Small discrepancies for Mercury, on the other hand, could only be explained after Einstein’s theory of relativity had overtaken Newtonian physics as a more complete description of the universe. The clear parallels in climate science have been critical examination of apparent anomalies in global temperature data (compared to model predictions), together with the impact of unpredictable events such as particular large volcanic eruptions. The normal scientific process of seeking increased understanding has, in this context, absurdly been characterised as data fraud, as part of the politicisation of the policy debates. Perhaps fortunately for Einstein and the general theory of relativity, views on the orbit of Uranus did not disturb as many vested interests.

Science may never be complete, but it is not always provisional. The absence of complete knowledge and understanding does not imply the absence of useful knowledge and understanding. Most of conventional science is for all practical purposes providing facts that we can and do treat as certain. Our understanding, even of the principles of quantum physics, and however incomplete, enables most of the advanced technologies of the modern world that we take for granted.

Popper’s characterisation of scientific knowledge as always provisional, is more relevant to particular parts of physics (quantum mechanics or astrophysics) than it is to many other sciences.  Theoretical physics does indeed include speculative interpretations, such as string theory or parallel universes (though it is questionable whether either is capable of being tested and meeting Popper’s falsifiability test). But much of science, especially in relation to the natural world, is descriptive and observational.

Science gives us a huge amount of fact that is certain. There are many more things that we “know” with certainty as a result of scientific observation, that do not need to be regarded as “provisional”, and are fundamental to any sensible consideration of public policy. This is true of much of climate science, which will be essentially concerned with careful observation of known phenomena such as parts of the carbon cycle or heat exchange within the oceans and the atmosphere.

 “Science is never settled. History tells us that!

This is an absurd generalisation and history tell us nothing of the kind. Science provides vast amounts of knowledge that we can and do, for all practical purposes, treat as certainty. We know that the basic theory of tectonic plates is true. We know that AIDS is caused by the HIV virus, and malaria by mosquitoes. We know that mass vaccination can protect populations from lethal diseases.

The anti-science brigade, from the Trump and Pence core vote on measles, or tobacco companies denying the links between smoking and lung cancer, or Melanie Phillips on MMR vaccines, or African leaders in denial on HIV/AIDs, and more recently anti-vaccination activists on Ebola, must over the years bear responsibility for huge amounts of human suffering.

The basic building blocks of climate science, in terms of the radiative forcing effects of various greenhouse gases, have well established parameters, and we have, at a minimum, an awareness of most of the other major factors involved in the carbon cycle and the climate system. And the knowledge is constantly improving. Climate science as a distinct discipline consists of the application of knowledge from a large number of separate but related fields of knowledge and observation. Putting the pieces together to form a full understanding of climate variation is a complex application of known science along with some known uncertainties. And of course it necessarily includes numerous year-on-year effects such as variations in solar variation and climate cycles such as el Nino, which do not have a human cause, as well as the possibility of sampling error in climate measurement.

 “We can’t even find models that forecast tomorrow’s weather, so there cannot be any basis for predicting the climate decades ahead.”

This is a popular myth with no basis in reality. Today’s short term weather forecasting, along with associated probabilities, is good enough to be extremely useful. In farming and shipping, not to mention storm alerts, it is regarded as a crucially important service. Forecasts will always have a margin of error, and will always be revised to reflect additional information and unforeseen extraneous factors, but in various forms it is an essential part of almost every human endeavour. 

The reality is that scientists have also proved remarkably consistent in their assessment of the determinants of climate, which is necessarily measured over much longer time scales than weather. This includes the global temperature impacts of greenhouse gases. These forecasting efforts are very well summarised in a Carbon Brief article that starts with some relatively simple calculations in the early 1970s.

In a paper published in Nature in 1972, Sawyer hypothesised that atmospheric CO2 would increase by 25% and that the world would warm 0.6oC between 1969 and 2000. Sawyer argued for a climate sensitivity – how much long-term warming will occur per doubling of atmospheric CO2 levels – of 2.4oC, which is not too far off the best estimate of 3oC used by the Intergovernmental Panel on Climate Change (IPCC) today, on the basis of much more information and sophisticated analysis.

In some instances, the forecasts have under or overestimated current observed warming because they were based on assumptions that under or over overestimated key parameters such as CO2 emissions, and other factors such as unusual volcanic activity. But of course the real purpose of forecasts is not absolute precision but a reasonably accurate picture, including the risks and uncertainties, of the impact of policy alternatives, such as the effect of allowing uncontrolled emissions or adopting measures for their limitation.

Rejection of known science is both foolish and unjustified. It has also become perhaps the most dangerous intellectual fallacy of our age.

[1] Popper had never read Principia Mathematica. Had he done so he might have found Newton’s ideas on science closer to his own.
[2] Science in the Looking Glass: What Do Scientists Really Know? E. Brian Davies Oxford University Press, 2003. 288 pages, ISBN 0198525435. (Useful review in Notices of the American Mathematical Society.) Inter alia this remarkable book provides a good description of Popperian ideas on falsifiability, both their positive contribution and their limitations.
[3] Climate Science and Falsifiability. Philosophy Now, 2014. Richard Lawson shows how Karl Popper can help settle the climate debate. This article by Richard Lawson provides a usefully brief discussion and turns the sceptic argument on its head by imposing a falsifiability test on the sceptic position.

[4] Christopher Booker. Trump and Pence.
[5] It is of course only in very specific conditions, such as astronomical observation or highly controlled laboratory experiments, that truly precise predictions and measurements are relevant. In most practical circumstances outcomes can be significantly affected by complex boundary conditions, eg the precise shape of the land mass and its contours.

No comments: