This post by Ben Goldacre
, brilliantly entitled 'Parmageddon', makes a lot of very good points about the coverage of swine flu, and the difficulty of reporting on a subject with a high margin of error. Without wanting to detract from that, I think it also misses a couple of important points.
The first is that the reason
that the public mistrusts newspapers on stories like this (Ben Goldacre lists Sars, bird flu and MMR together, which I think is an interesting grouping, and I'll explain why later) is that the headline writers - and usually the copy writers too - instinctively go for the most extreme end of the margin of error. Faced with a study that says that sharpening pencils is probably harmless, but there's a 2% chance that it will cure cancer, and a 1% chance that it will attract meteorites, the headlines will be 'Flaming fireballs to rain destruction on school classrooms' one day, and then (for balance) 'Put lead in your pencil for a long life, say scientists' the next. Nowhere is there any coverage of the outcome that is vastly more likely, because that doesn't sell newspapers; to be fair, it doesn't sell newspapers because people find it boring to read, so it's not completely
the journalists' and editors' fault.
Then along comes swine flu - a phenomenon where, as Ben Goldacre points out, it's really too early to say at this stage what the likely outcome is. We don't have enough data to make an educated guess. We can do the Bayesian thing
and attempt to assign probabilities to hypothetical outcomes, but I don't think (and AFAICT Ben Goldacre agrees) that we can do it with any degree of confidence. So when the papers say '40% of the world could be infected' or '120 million people could die', it's a Bayesian 'could', and one that we don't really have a good feel for yet. We can look at past flu epidemics (although the problem here is the tendency to focus on the obvious, severe ones), but even so, we don't yet know whether this one's like any of those.
It's chaos theory
. We're looking at a tiny pressure difference in the mid-Atlantic, which might fizzle out and might develop into a hurricane, but we can't measure it in enough detail (or model its complex behaviour well enough) to say yet.
Of course, the World Health Organisation and Governments need to at least try to do these calculations, and the understandable tendency is to prepare for the worst scenario that looks at all likely. But even they, at this stage, are almost certainly dealing with judgements that they simply don't have the data to make accurately at this stage. And the media reports this, as it should, although often not with the emphasis it should.
As Ben Goldacre rightly points out, that's not to say they're wrong to make such judgements, nor that they will have been shown to have been wrong if the threat doesn't materialise. When the hypothetical 'this might happen' becomes an after-the-fact 'this did happen', we're prone to looking at every 'this might happen' that didn't
, and concluding that they couldn't
have. (We take a frequentist approach to a Bayesian analysis, in other words, forgetting that we couldn't do a frequentist analysis before the fact anyway.) Sars and bird flu didn't
become global threats (yet), but that's not to say that with slightly different starting conditions which would have been impossible for us to measure at the time
, they could never have done so.
And then there's MMR, and the possible link to autism. I find it odd that Ben Goldacre groups this to Sars and bird flu at the start of his article, because he of all people should know that this really was
largely a panic engineered unwittingly by the media, with no good evidence to back it up. Sars and bird flu both could have been serious threats; MMR couldn't have caused widespread autism, and this should have been apparent at the time. What's more, the fact that the story about the supposed MMR link to autism has almost certainly caused vast damage by reducing the takeup of the vaccine is a neat illustration of the fact that it's often the side effects of the story - the public's over-reaction to it - that you need to watch for. It may turn out to be that case with swine flu; if it does become a serious threat, I worry that far more people will be killed in a scrap over medical and food resources than will be killed by the flu itself. So far, though, this shows no signs of happening, which is a very good thing.
Ben Goldacre mentions the boy who cried 'wolf', and (correctly) points out that the public (incorrectly) see the headlines about swine flu to be an example of this. This is true of the story as told be Aesop; in the fable, the motivation of the boy who cried 'wolf' was to get attention for himself. However, if the boy cried 'wolf' because he thought that the thing he saw moving in the woods near the flock was indeed a wolf, even though he knew full well that it could also be a deer, that's a different story in more ways than the obvious one. Is he then wrong to cry 'wolf', even if it turns out to be a deer? Are the villagers right to ignore him the third time, after it's been a deer for the first two, knowing that there are real wolves out there? Would the village newspaper be scaremongering to print 'Wolf may lurk in wood'? Possibly, but it's a lot less clear-cut.
(Cross-posted from Dreamwidth here: http://djm4.dreamwidth.org/644.html
- feel free to comment in either place. I promise my LJ isn't only
going to be Dreamwidth cross-posts from now on.)