It is a popular and longstanding fallacy that scientific knowledge grows by inferring conclusions from data. The notion goes back to Bacon, who felt that the way to do science was to cast aside preconceived ideas, make observations and think carefully about how to understand them. Thus, reason would lead to knowledge unimpaired by our preconceptions about how the world worked. However, thinking about phenomena without a theoretical framework from within which to understand them is extraordinarily difficult. It is easy to see the problem from a modern perspective. For example, consider the Stern-Gerlach experiment, which famously demonstrated that angular momentum is quantised. Silver was heated, and gaseous silver atoms were formed into a beam that passed between the poles of a magnet to strike a screen. Two distinct spots were observed on the screen. But how can we infer that these spots tell us something about angular momentum? We need some theoretical understanding of the interaction between a magnetic moment and a magnetic field, and a hypothesis that something exists in the atom (the electron spin) that is capable of having two distinct values (up and down) that interact with the magnetic field in the manner of objects that have magnetic moments. The observations in this famous experiment are, in the words of philosophers, theory-laden, as is most of modern science.

In the early days of natural philosophy, the notion of “inductive reasoning” gained widespread currency. The Baconian picture of science as a process based upon careful observation followed by inductive reasoning from observation to theory became, for many, synonymous with the scientific method. However, a problem lies at the heart of the Baconian model. Observations are necessarily finite in nature; we can only make a limited number of observations because human beings have limited capacities. But our theories claim much broader scope – indeed, modern science claims universal validity. Thus, cosmologists seek both to explain the past and to predict the future of the universe based upon theories that are supported by finite measurements made during a short period in the universe’s history with instruments that are all subject to limitations of some sort.

The British philosopher David Hume realised that there was a fundamental problem with inductive reasoning. In trying to understand the world in a mechanistic way, we might identify certain things as causes and others as effects. But how can we reason from a finite number of observations to a law that has predictive power? Even where a cause and an effect have, in our experience, always been related in a certain way, how can we guarantee that they will be so related in the future?

“Thus not only our reason fails us in the discovery of the ultimate connexion of causes and effects, but even after experience has informed us of their constant conjunction, ‘tis impossible for us to satisfy ourselves by our reason, why we shou’d extend that experience beyond those particular instances, which have fallen under our observation”.

David Hume, A Treatise of Human Nature, Book I § VII

Hume’s analysis exposed a fatal flaw in the Baconian conception of the scientific method. It was two hundred years before a resolution emerged in the work of Karl Popper. For Hume, the problem was to find a basis for discovering the connection between causes and effects, but Popper turned the problem on its head; instead of attempting to reason from finite observations to universal theories, he said, we should reason deductively from hypotheses with claims of general validity to finite tests. The defining characteristic of scientific hypotheses was their testability, not their freedom from preconceptions. Scientific hypotheses assert connections between causes and effects but we do not justify these hypotheses inductively, by developing a connecting chain of reason from cause to effect; rather, their scientific status consists in our ability to reason decuctively from the hypothesis to a test. Indeed, Popper stated very clearly that he was not interested in where hypotheses came from; the creative process was psychological and unscientific. What he was concerned with was the justification of those hypotheses once they had been formulated.

Consider a very trivial scientific law: “all swans are white”. To prove the truth of this law, one must observe all the swans that exist, that have ever existed and that ever will exist. Such a proof of truth is impossible; based on a finite number of observations of swans, “‘tis impossible for us to satisfy ourselves by our reason, why we shou’d extend that experience beyond those particular instances, which have fallen under our observation”. Popper relieves us of this impossible burden of proof: we begin with a universal statement (“all swans are white”) but assert that this hypothesis will be falsified on the observation of a single black swan. Thus, the hypothesis is testable (i.e. scientific), and the test is finite and founded on deductive, not inductive, reasoning.

For Popper, to think scientifically is to articulate testable hypotheses about the world. When these hypotheses are tested, they either pass the test (they are corroborated – never proven to be true) or they are falsified. A falsified theory is replaced by a better theory, with greater explanatory content, and knowledge grows. For example, Newtonian mechanics fails to describe the behaviour of electrons in atoms, but it is not necessarily “wrong”: it helps us to design aricraft that we use very happily to travel to foreign lands. Quantum theory “falsified” Newtonian theory in the sense that it explained more, and solved problems that classical physics failed to explain.

The very important conclusion from this short summary of Popperian thinking is that above all else, the justification of scientific knowledge is based on a method: a hypothesis is conjectured, a test is established by deductive reasoning, the test is completed and the hypothesis is either falsified or corroborated. Thus, science does not consist in the statements of scientists, but in the set of testable hypotheses about the nature of reality. Statements made by scientists that are not testable are not scientific.

As the Covid-19 pandemic wore on, interviewers on TV and radio began to ask scientists “when will we return to normal?” There are a number of scientific responses to this. One is that eventually, Covid-19 may become endemic. In the mouths of Government ministers, “endemic” describes sunlit uplands in which nobody has to worry about infectious disease. However, in epidemiology, a disease can be considered endemic when it has reached a steady state. Malaria is endemic in much of the world, but that provides no comfort to the many who fall victim to it every year; endemicity has not rendered the mosquito’s bite harmless.

Endemic is a scientific term with a scientific meaning that is testable in relation to any particular disease according to Popperian methodology. For a reproduction rate R₀ and a susceptible population of size S, a disease is said to be endemic when

R₀ x S = 1

This is testable in principle for Covid-19, although determining the magnitudes of both of the variables in this simple equation is a non-trivial undertaking.

But many scientists have gone much further when interviewed by the media. For example, Paul Hunter, a biologist from UEA is quoted as follows by the Guardian:

“This is a disease that’s not going away. Ultimately, we’re going to have to let people who are positive with Covid go about their normal lives as they would do with any other cold,” he told BBC Breakfast. “If the self-isolation rules are what’s making the pain associated with Covid, then we need to do that perhaps sooner rather than later. Maybe not quite just yet.

“Covid is only one virus of a family of coronaviruses, and the other coronaviruses throw off new variants typically every year or so, and that’s almost certainly what’s going to happen with Covid. It will become effectively just another cause of the common cold.”

The argument goes that one cause of the common cold (not the main one – most colds are caused by rhinoviruses) is a harmless coronavirus which, when it first emerged – as the “Russian flu” in 1889/90 – was a more harmful disease. Over time, it has evolved to become comparatively harmless. Hence, Professor Hunter infers, the same fate will befall Covid-19. Other sientists, including epidemiologists and viroologists, have said very similar things. Coming from a scientist, these sorts of statements seem to be incredibly reassuring, and in the mind of the public, science is synonymous with the statements of scientists, whereas according to Popper – as we have seen – it consists in a set of testable statements about reality.

In fact, it is not clear that the development of Russian flu, caused by human coronavirus OC43, is quite so straightforward as all that. Belgian virologist Marc Van Ranst suggests that while OC43 is a cause of the common cold, it also tends to cause more severe illness, and may account for significant mortality among elderly patients to this day. So the first thing to note is that there is a need for caution in summarising virological evolution for the benefit of the general public; for over 100 years, OC43 spread without genetic surveillance, so reconstructing its impact over that time, and the resulting mortality, is very difficult. Moreover, as Van Ranst notes, the pathways for viral transmission are dramatically different now compared to 1890.

But that is not really my main point. Paul Hunter’s statements and those of other scientists seeking (for the best of reasons) to offer reassurance to a public weary of Covid-19 are a very up-to-date illustration of the problem of induction. Hunter says that a previous coronavirus evolved to become less harmful (the extent to which that statement is corroborated by data is unclear) and that we can assume that the same fate will befall Covid-19. In other words, based on a single observation of cause (the passage of time) and effect (that OC43 became to some extent less severe) a generalisation is made that this will also be the outcome for Covid-19. One could argue that this constitutes a scientific hypothesis: we need to wait and see. However, I am not aware that any of those offering this optimistic hypothesis are defining under what conditions the hypothesis would be falsified. Rather, it looks much more like the naive application of inductive reasoning.

If Covid-19 has disappeared in six months’ time, then of course those favouring the belief (I think it can be fairly described as a belief, because the evidential basis is flimsy) that Covid-19 will necessarily evolve into a less severe illness may be able to claim it has been corroborated, but what if the pandemic is still raging in a year’s time, or two years’ time? Compared to Newton’s laws of motion, that are falsifiable but which have been tested hundreds of thousands, probably millions of times and been largely corroborated, the degree of confidence that we can reasonably have in such an hypothesis is very limited indeed.

This brings us back to public policy. Those who prescribe pathways for managing infectious diseases (the WHO and the Government’s Chief Medical Adviser) should not develop policy to conduct scientific experiments, because lives are at stake. The ethical approvals process for experiments involving human subjects is rightly onerous and exacting; public health policy relies instead upon what has been found to work in the past, and on hypotheses that are corroborated by scientific testing. Public health decisions should not be based upon clever and creative punts that are untested by scientfic scrutiny and testing (or worse, untestable). Many voices – besides those in the Government – appear to be encouraging the public to look towards the sunlit uplands of endemicity. However, what has been almost completely lacking from the public debate, or from the statements of scientific experts, has been a detailed and objective evaluation of what “endemic” might mean for Covid-19. Will Covid-19 evolve to become a mild disease, or will we need to implement mitigations for decades to come? Only time will tell. There are, to my knowledge, no testable and tested scientifc hypotheses that can inform us about what the future holds. However, thinking about the future in a rigorous and properly scientific way can help us to prepare for whatever outcome lies in store.