Science Shorts 1: Facts and Values

Facts and values

When it comes to making statements about the world, we can consider two types of claims. “Is” (or factual) claims concern the way the world was, or is, or will be. “Ought” (or value or normative) claims concern the way the way the world should have been or should be.Footnote 1

For decision-makers, the distinction between facts and values is important for at least two reasons. First, science is directly concerned with factual claims: it is not directly concerned with normative claims. To the extent that decisions are informed by factual claims, science may provide evidence concerning the truth (or not) of such claims. But science cannot directly inform questions about what values are most important to respect or consider in making decisions.

Second, inferring the validity of a normative claim solely from the truth of a factual claim is problematic.Footnote 2 Suppose, for example, that wearing masks does indeed reduce the risk of SARS-CoV-2 infection. Even if this factual claim is true, one cannot infer that therefore we should have mask mandates. Why? Because for all decisions, there are multiple values to consider. Although reducing the rate of spread of COVID-19 is an important policy objective, so too might be minimizing limitations on personal choice or minimizing the amount of personal protective equipment waste. If either of these last two objectives was considered more important than reducing the spread of COVID-19, then we would be unlikely to impose mask mandates even if wearing masks does reduce infection risk.

“Following” the science

Science cannot directly tell us what we should (or shouldn’t) do: it can only tell us what is likely to happen if we do (or don’t do) something. For example, no science in the world can tell us we should have vaccine mandates. But medical science can potentially tell us what the effects of a vaccine mandate on, say, the rate of spread of COVID-19 are likely to be. And economic science can potentially tell us what the effects of a vaccine mandate on, say, employment in the service industry are likely to be. Both estimates may be useful for informing decisions about vaccine mandates.

Desired and undesired outcomes of policy decisions are simply expressions of the values considered to be most important to a decision. Once these outcomes have been specified, the relevant science (a) identifies how desired outcomes might best be achieved or undesired outcomes avoided; or (b) informs the likelihood of achieving desired outcomes, or avoiding undesired outcomes, if a specific decision is taken. “Following the science” then amounts to a decision for which there is sufficiently compelling scientific evidence that the decision not only can achieve the desired outcomes (or avoid undesired outcomes), but indeed will do so.

Values in science

Scientists strive to be objective, unbiased and impartial. But because scientists are people too, Olympian objectivity and impartiality is impossible.Footnote 3 For this reason, decision-makers must consider the possibility that scientific results — or their interpretation — may be influenced by competing interests,Footnote 4 conscious or unconscious biases,Footnote 5 social influences or even the scientist’s own values.Footnote 6

Even science itself is not value-free.Footnote 7 For example, in deciding whether an hypothesis is supported, scientists must decide whether study results match predictions. Since there is never a perfect match, this means deciding whether the match is “close enough” to legitimately infer that the study results support the hypothesis.

“How close is close enough?” is not a factual question, it is a normative question. Although scientists have (more or less) agreed on a consensual answer, this answer reflects a fundamental value claim: for scientists, the error associated with inferring that an hypothesis is true when it is actually false is worse than the error of inferring that it is false when it is actually true.

Kinds of facts

There are four kinds of factual claims: observations, estimates, patterns and causal hypotheses – facts of the first, second, third and fourth kind respectively (Table 1).

Fact type Description Example
Observation/measurement (1st kind) A claim about the value of an individual observation or measurement. The temperature in Ottawa at noon on Feb. 16, 2022 was -14°C.
Estimate (2nd kind) A claim about the value of some quantity based on a collection of individual observations or measurements. The average individual Canadian income in 2019 was $49,000.Footnote 8
Pattern (3rd kind) A claim about the empirical relationship between two or more attributes of interest from a collection (“sample”) of observations. In response to vaccination, concentrations of neutralizing antibodies peak at two weeks post-injection.Footnote 9
Causal hypothesis (4th kind) A claim about the effect of a change in one attribute of interest (the “cause”) on one or more other attributes of interest (the “effect”). Physical distancing reduces rate of spread of COVID-19.Footnote 10

The distinction among different kinds of facts is important for decision-makers for at least two reasons. As one moves from factual claims of the first to the fourth kind:

  • The difficulty of establishing the truth of the claim increases dramatically: it is far easier to determine that the temperature is -14°C than it is to determine that self-isolation after a positive PCR test reduces the rate of spread of COVID-19.
  • The risks to one’s values often increases dramatically. Whatever my values are, whether it is indeed -14°C outside is unlikely to substantially affect them. But whether the claim that self-isolation upon testing positive reduces spread of COVID-19 is true may have profound implications to many of my values — and to yours.

For these two reasons, facts of the third — and especially of the fourth — kind are much more likely to be contested than are those of the first or second kind, not only by the public, but by scientists themselves.

Predictions about the consequences of alternative decisions are always based on one or more (usually implicit) underlying causal hypotheses — precisely the kinds of facts whose truth is the most difficult to establish and which are the most contested. It should, therefore, come as no surprise that even when the best available evidence is used appropriately, decisions affecting important values will always be highly contested, and actual outcomes may well not match those predicted.