Could popular science do a better job on uncertainty?

I think that popular science does not convey the ambiguities and uncertainties of real science. Science has plenty of ambiguities. The deeper I go, the more ambiguous it feels. Scientific ideas span the gamut from extremely certain to hunches. Some things are in between; we might call them “reasonable hypotheses.” There is a distinction to be made between a reasonable extrapolation from existing data and conclusions directly derived from data.

Here’s an example: “dietary fat causes people to get fat.” That’s a hypothesis with lots of reasonable justification. But it was presented as an empirically derived fact. It was presented as if this premise was borne out by lots of data. Recent years have suggested otherwise.

When people eat more calories than they expend they make fat. That is an empirically derived fact. Fat has more calories per gram than other foods. That is an empirically derived fact. The association between high-fat diets and weight gain is a reasonable supposition based on these two facts. It is not an empirically derived fact. The data does not support that conclusion. A meta-analysis of more than 3000 studies concluded: “In weight loss trials, low-carbohydrate interventions led to significantly greater weight loss than did low-fat interventions… Low-fat interventions did not lead to differences in weight change compared with other higher-fat weight loss interventions.”

People might feel like they had been lied to. It wasn’t a lie: it seemed reasonable at the time. “Eat fewer calories by cutting fat” seemed like good advice. It just wasn’t. And now the data is in. We can change behavior. Hopefully. Unless people don’t want to trust the advice on account of the fact that they now distrust the whole enterprise.

Conflating fact and extrapolation is dangerous. Treating a reasonable hypothesis like a fact can, ultimately, degrade the credibility of science.

Treating facts as if they were hypotheses is just as bad. The Merchants of Doubt documentary talked about this at length. It’s easy to claim anything is “just a hypothesis.”

At one point, someone hypothesized (reasonably) that since tobacco smoke has nicotine and carbon monoxide and tar, it might be harmful.  It started as a just a hypothesis, not a fact.

The tobacconists don’t want to think that they are selling addictive poison. So they start off by saying “the data is not yet in, we can’t be sure.” At first, this is true. After more and more data gets collected, the poison hypothesis moves into fact territory. But it’s easy to move the goalpost and keep insisting that it’s still a hypothesis. Those arguments still sound valid.

I’m not sure how to fight that, but I suspect it would help if we more often pointed out which is which.  But that’s just a hypothesis.