HOW probable is probable? When we say something is unlikely to happen, what
do we mean? 杏吧原创s and science writers alike are notoriously bad at
communicating a degree of uncertainty to the world at large.
For instance, when scientists said British beef was safe to eat, they did not
mean absolutely safe, only relatively safe. The nation was astounded. This is
simply not good enough. New 杏吧原创 has had similar difficulties. In
a recent issue, no fewer than 17 of the 47 news stories included the dreaded
word 鈥渃ould鈥� in the first sentence. For example 鈥淲ithin two or three years, a
simple blood test could tell you whether . . . 鈥� But how likely is it? Bookies
have cracked the problem of getting over the rules of chance鈥攕o why
haven鈥檛 scientists? Up to their ears鈥攅ven floundering鈥攊n
uncertainty, are the boffins on the UN鈥檚 Intergovernmental Panel on Climate
Change. Three years ago they announced 鈥渢he balance of probability鈥� indicated
humans had caused recent global warming. I don鈥檛 doubt their integrity, and have
few fears about their conclusion. But when pressed, they say this balance of
probability amounted to little more than the result of a straw poll among
themselves.
But no longer. The IPCC scientists are smartening up their act and bringing a
little statistical order to the chaos. Marshalling them in this task is Stephen
Schneider, a climate modeller at Stanford University in California who, as a
young man, outraged his fellow white-coats by voicing his radical views on
global warming. Now all this has been turned on its head. Those radical ideas
are now mainstream, while Schneider is counselling caution. And he is right.
Advertisement
In a recent internal IPCC paper, Schneider called for 鈥渃onsistent assessment
and reporting of uncertainties鈥�. No more unquantified 鈥渃oulds鈥�, no more straw
polls. Instead, scientists should prepare a 鈥渢raceable account鈥� of how estimates
are constructed. Perhaps, like me, he recalls past IPCC reports that blandly
passed on spurious-sounding predictions, 鈥渟cenarios鈥� and 鈥減rojections鈥�. To take
a couple at random, I remember being told to expect a 23 per cent decrease in
salmon in the Sacramento River and a 43 per cent drop in the flow of the River
Indus.
Schneider鈥檚 pleas could be read as a manifesto for conservatism. But at root,
he is calling for more honesty, and that could revive radicalism by giving it
greater clarity. His paper warns that there is a 鈥渨ell-documented tendency鈥� for
scientific committees to overstate the confidence of their guesstimates. IPCC
teams must resist this and go out of their way to identify extreme values at the
top and bottom end of a range of possible outcomes, he says. He goes on to note:
鈥淪ome authors are likely to feel uncomfortable with the full range of
uncertainty because the likelihood of a `surprise鈥� . . . may be extremely remote
or essentially impossible to gauge.鈥� But dammit, that is their job.
And this brings us back to my original question: how probable is probable?
Here Schneider has no difficulty. In his new classification of probability,
Schneider defines 鈥減robable鈥� as a two-thirds chance. Warming to his theme, he
rules that from now on, when IPCC scientists say they are 鈥渃onfident鈥� of
something, they should mean they are nine-tenths sure. And 鈥渧irtually certain鈥�
will up the stakes to 99 out of 100. At the other end of the scale,
鈥渜uestionable鈥� means 3:1 against, 鈥渦nlikely鈥� 10:1 against and 鈥渧ery unlikely鈥�
100:1 against.
Some people will rail against the rigid precision of these definitions. But
the irony is that it should free researchers to speak more openly about all the
鈥渦nlikely鈥� climatological nightmares鈥攖hose hundred-to-one shots, some of
which, probability being what it is, will undoubtedly happen.