SCIENTISTS ON TAP, NOT ON TOP

In the summer of 2021, the UK’s Joint Committee on Vaccination and Immunisation (the JCVI) was asked whether the COVID-19 vaccination programme should be extended to adolescents. Their response was complicated, but, ultimately, a ‘no’. This recommendation created a political headache, with scientists and commentators loudly demanding the UK follow other countries in vaccinating adolescents. So, the government sought a second opinion—from the Chief Medical Officer—and, on this basis, went ahead with vaccinating teens. Unsurprisingly, vaccine sceptics made much of this mess: the government had justified its often draconian COVID-19 policies as ‘following the science’, but here they were disregarding the scientific advice.

These criticisms were unconvincing but in the sceptics’ defence, ‘following the science’ was a ridiculous slogan. Scientists have an important role to play in advising policymakers. However, ultimately, in a democratic society, it is up to elected officials to decide what to do. In Winston Churchill’s phrase, advisors should be ‘on tap, not on top’. If so, it can’t be true that policymakers should always do precisely what the scientists recommend. Otherwise, we might as well have a full-blown technocracy. I think these platitudes are correct. Unfortunately, they just raise a further question: what then is the proper role of scientists in policy?

In the early twentieth century, the great German sociologist Max Weber sketched a general account of the relationship between scientific experts and policymakers. To simplify, Weber’s work rested on three claims. First, all policy-making involves two kinds of inputs: factual claims (which help us identify effective means to achieve our goals) and evaluative claims (about what goals to pursue). Second, there can be ‘experts’ on factual claims. Third, there are no ‘experts’ on evaluative claims.

Clearly, if there are experts on factual claims, they ought to provide the factual inputs to policy. How, though, should we decide on the evaluative claims? Weber had a controversial answer to this question in terms of charismatic leaders. By contrast, I assume that the evaluative inputs to policy should depend on what the public do (or would) choose through a democratic procedure.

So, policymakers should decide which policy to follow by drawing on factual inputs from scientists and evaluative inputs from the people. But how do we ensure scientists remain on tap, not on top?

Here’s one option: If we decide not to vaccinate adolescents, that might because we have assumed that it is more important to protect teenagers from side-effects of vaccination than their grandparents from COVID-19. That is to say, any time we make a normative recommendation (we should do something), we implicitly assume some evaluative claim (this is what is important). So, if scientists make these kinds of recommendations, this looks as though they are going beyond their proper remit. Therefore, the argument runs, scientists ought to limit themselves to making factual claims. I call this the ‘just the facts’ option. Unfortunately, it is both too strong and too weak.

Why is it too strong? It is often very hard to understand how technical scientific findings relate to policy. Therefore, when scientists’ evaluative assumptions (‘this is what is important’) are aligned with democratic values, it is helpful for them to make explicit recommendations. ‘Just the facts’ rules out some good practice.

Why is it too weak? Whenever we list facts, we must choose what to say and what to leave out. The alternative is a data dump, which doesn’t help anyone. So, even when scientists list facts, they make implicit judgements about which of those facts are relevant given what they think is important. ‘Just the facts’ doesn’t rule out enough.

With these concerns in mind, consider a more plausible way to describe things: Whenever scientists research policy-relevant topics or advise policymakers, it is ‘as if’ they make assumptions about evaluative issues. So, for example, if a scientist collects data on the effects of vaccinations on public fear, then communicates this information to policymakers, it is ‘as if’ she is saying that this is a relevant consideration in the giant policy calculus. Clearly, we cannot avoid scientists making these ‘as if’ judgements. This looks like a problem for using science. However, it need not be, as long as those ‘as if’ choices are aligned with democratic values. If the public think reducing public fear is a valuable goal, the scientist ought to report on this topic; if they do not, she should not. I call this the ‘scientific alignment’ option.

A version of this scientific alignment proposal is implicit in much recent writing on the relationship between science and policy, and for good reason: it seems very attractive. In particular, it promises a response to recent work arguing that all stages of the scientific research process—not just the communication of results—are ‘value-laden’. On the face of it, claims that scientific research is value-laden challenge the use of science in democratic policymaking, because they imply that scientists ‘vote twice’: once as scientists and again as citizens. The alignment proposal seems to avoid this problem: ‘sure, science is value-laden, but only with the democratically agreed values’.

But even if scientific alignment is a wonderful ideal, it is unattainable. To see why, let’s take a break from science and policy to think about sandwiches. Imagine you are given a sandwich. You realize it contains cheese, but you prefer egg. It’s tricky, but you can remove the cheese, replace it with egg and still have something edible. My worry is that for scientific alignment to be possible, values in science must be like the sandwich filling: easily identifiable and replaceable, while maintaining integrity. Unfortunately, they aren’t.

Our JCVI case illustrates this problem. The committee’s recommendations against vaccinating adolescents were controversial in at least three ways. First, in assessing costs and benefits of vaccination they focused solely on medical outcomes, ignoring, for example, effects of missed schooling. Second, they focused solely on outcomes for adolescents, ignoring the possible effects of vaccination on transmission to more vulnerable groups. Third, and strangest, even after limiting focus to this narrow range of outcomes, the JCVI conceded vaccination was probably in adolescents’ interests, but recommended against on ‘precautionary’ grounds.

What was going on? The recommendations were deeply shaped by a commitment to a norm of medical ethics: ‘do no harm’. But this norm can conflict with more utilitarian ethical approaches, which aim at doing as much good as possible. Plausibly, the JCVI’s advice assumed a strong preference for a ‘do no harm to the adolescent’ approach, whereas the government was willing to adopt a utilitarian ‘do the most good for the population’ perspective. The values of the advisors and the government were misaligned.

If that were the end of the JCVI story, the sandwich-filling model would be right: we can identify the JCVI’s evaluative assumptions, figure out some replacements, and get some alternative advice. Alas, the situation was more complex again. When we dig into the JCVI’s decision-making, they justify silence on various issues—say, the effects of missed school on educational outcomes—on the grounds that there is no ‘high-quality’ evidence. They frame their judgements as scientific, rather than ethical.

However, this framing, while undoubtedly sincere, is problematic if we are concerned about misalignments. First, while it may well be true that some data was shabby, it’s not obvious that we should therefore simply ignore that data for practical decision-making. Indeed, to do so is to assume that it is more important to be cautious—to do no harm—than to try to do as much good as possible. Second, the JCVI’s claims about what counts as high-quality data rested on the claims of the ‘evidence-based medicine’ movement. This movement itself was motivated by a strong ethical concern that other approaches were breaching the ‘do no harm’ norm, and this conviction helped to shape the movement’s methods.

There is no easy way of disentangling the JCVI’s scientific practices from evaluative assumptions. We cannot simply remove and replace a commitment to ‘do no harm’ with ‘do the most good’. The entire research field was arranged to prioritize the former over the latter. Values in science are not like the cheese in the sandwich filling, but more like the yeast in the dough.

If the ‘just the facts’ approach and scientific alignment have both failed to capture what’s going on, what should we say instead?

Here is the proposal I lay out in my BJPS article: We should welcome misalignments! We shouldn’t have a single set of experts who advise policymakers. Rather, we should welcome a wide range of different expert groups, each providing a recommendation from a different perspective. In this way, we recognize the fact that scientific claims are always inherently ‘as if’ value-laden, while also minimizing the risk that science-based policy implicitly assumes the wrong, non-democratic values.

This proposal strikes many people as problematic. They worry that it will lead to politicians cherry-picking scientific advice on the basis of narrow electoral goals rather than the common good, or that it will promote a cacophony of competing experts promoting highly dubious ‘alternative facts’. These are, I admit, real problems and concerns. Still, really, they are problems with our political system more broadly and the incentives it creates for mischief. We can’t bypass them by pretending that science advice can be value free, as with the ‘just the facts’ proposal, or that values in science function like sandwich fillings, as with the scientific alignment proposal.

More positively, consider two advantages of this proposal. First, it matches everyday advice. If your doctor recommends that you have an operation, but your bank manager recommends that you do not, it is entirely possible that both pieces of advice are correct: from a medical perspective, this is a good idea; from a financial perspective, it is not. Nothing much would be gained by having the doctor and bank manager form a super-committee. Rather, as in the ancient Buddhist legend of the blind men each feeling a different part of an elephant, you need to build up a whole picture of the situation from multiple partial reports. The notion of differently located experts providing multiple perspectives is a very familiar feature of our lives.

However, in making a decision of this sort, you need to do more than just build a picture: you also need to decide what you value more, health or finances. This leads to the second nice feature of my proposal: that it makes politicians honest. Weber himself wrote that politicians try to hide trade-offs, pretending that they have found policies that will promote all the good things at the same time. He suggested that one important critical role of the scientist was to call out this wishful thinking, to say we can promote this value or that value, but we cannot simultaneously promote both. Modern policymakers are much the same as in Weber’s time. They act as if a decision to vaccinate or not vaccinate adolescents will never involve trade-offs between different values, but is just dictated by the science. If we get rid of the myth of ‘the’ scientific advice and, instead, allow for many voices, then we are better placed to identify who wins and who loses from a policy.

‘Build Weber’s Elephant’ is a less catchy motto than ‘follow the science’, but a better guide to the messy fact that science alone can never resolve our questions.