The “Science” is solid
At the risk of becoming a climate change bore, once more into the breach.
A meme bouncing around about the 4th report of the IPCC is that “the science is solid“. One of my reading hobbies is science, mainly physics and philosophy of science. I’ve been reading it for years and thinking about the nature of science as an enterprise.
The main thing which distinguishes science from most other human activities is a process of hypothesis and experiment. Basically, science advances because a particular scientist develops a view of the world which makes a specific and testable claim about the world. The claim is tested and, if it is proven true, is added to the store of things we know - scientifically - about the world. Evidence accumulates for the theory which generates the hypothesis.
The key thing here is not “consensus” which is social rather than a scientific concept. Rather it is the idea that a hypothesis needs to be tested experimentally to determine its truth or falsity.
Now, we are going to have to wait for several months for the scientific papers which underlie the IPCC report to be made available; however, reading that report does not reveal a single testable hypothesis or proposed experiment.
So, is this science? Models - and the IPCC to date has been all about models - are about building simulations which account for the data to date. However, the test of a model, is its capacity to make predictions which, in their turn, can be tested against future events.
My partner and I, when our first child was born, marveled at his capacity to gain weight. Birth 6.5 pounds, week 1, 6.8 pounds, week 2 7 pounds, week 3, 7.3 pounds, week 4 7.8 pounds. The kid was gaining half a pound a week. Now a straight line model of baby growth would have meant Sam would have weighed between 30 and 35 pounds at one and 60-70 pounds at two.
As it happens doctors chart such things and have done so for a very long time. For a male child in the 95th percentile of weight 27 pounds is the intercept at 12 months, for a male child in the 5th percentile the intercept is just under 19. So my straightline model would have been wrong by between 10 and 50% at one year. However, at two years the 95 percentile intercept is 33 1/2 pounds. Ooops.
And, indeed, Sam’s weight gain curved away from the straightline model. Which, had I been a scientist testing a hypothesis, would have refuted the straightline hypothesis for first and second year baby growth.
Obviously, real science is more complicated than this but the principle holds. What people normally mean when they say “the science is solid” is that the hypotheses generated by a particular theory about the world have been confirmed, to some extent, experimentally - or are at least consistent with the state of the experimental art in the particular field. This does not make the theory true; rather it makes the hypotheses it throws up confirmed by repeatable experiments.
Does the IPCC 4th report suggest any experiments which might confirm the hypotheses of man made global warming? Does it make specific near in predictions which might - absent actual experiments - confirm the models upon which it is based? And, in the now two decades worth of work done by the IPCC have there been predictive or experimental confirmations of its modeling? So far as I know - and I can, of course, be refuted - this has not been how the IPCC has proceeded.
Instead the IPCC and its enabling bureaucrats have proceeded by way of “consensus”. Consensus is an excellent way of picking a restaurant but it is a long way from science. The problem being that it is a political and social rather than scientific construct. The boring example here is the consensus which supported the 17th century geocentric view of the solar system and, indeed, Universe. There was agreement, models, data, authority and scholarship - “But yet it moves.”.
The construction of large scale computer models of a system as large and as overdetermined as the Earth’s climate is a remarkably difficult enterprise. Paleoclimatology, astronomy, atmospheric chemistry, oceanography, various branches of applied mathematics and physics all have their contributions to make. And, if you are proposing that humans are adding their mite to the mix, you will have to bring in demographers and economists and all manner of sociologists and anthropologists.
Realistically, you will have to construct several models of the Earth’s climate and run them on backsets of data. And you will then have to see how each performs against the data which Nature offers up everyday. And then, to convert this exercise to meaningful science, you will have to use your models to generate hypotheses about the Earth’s climate. Once you begin to do that you are actually doing science.
People make fun of the Kyotologists because they see every fluctuation in the weather as confirmation of the global warming theory. A warm summer…global warming; no winter…global warming; -30 in January…global warming. However, absent any clear, refutable, hypotheses offered up as hostages to fortune by the climate change consensus, it is impossible to see what would count as evidence for man made global warming or what might count against it.
There are many reasons not to buy into the IPCC’s current report and one of the most basic is that it does not meet the basic test for being science.
Special for GeeksI was thinking about the above and realized my baby example does not quite get it. Here’s another, admittedly social science example.
Propose that I had a theory that when times were good incumbents were re-elected. My hypothesis would be that a robust economy produced a statistically significant increase in the number of incumbents elected. To test that hypothesis I would have to build a model which a) contained a variable or variables going to rates of re-election for incumbents, b) contained a variable or variables to measure the robustness of the economy.
To help test the model I would run it using backdata: that is the actual values for the variables I was testing and, when I had a model which seemed to fit the data (and, yes, I do know that is the wrong way to make a model in real life) I would then run it on current data and make a prediction as to the rate at which the incumbents would be returned. My hypothesis would be that the stronger the economy the greater the likelihood of incumbent re-election all other things being equal.
Now, if I ran it on a data set where the economy was strong and a statistically significant number or incumbants were re-elected I could have some confidence in my model. Repeated correct predictions would increase my confidence.
The more interesting case, however, is where the hypothesis fails. More interesting because this could be the result of a) my hypothesis being wrong, b) my model failing to take into account some variable or set of variables which it should have.
The problem with modeling is that it is difficult, if not impossible, to be certain that your model is, in fact, a realistic and sensitive representation of the actual world. Variable selection and weighting make a huge difference to the final output. Spurious correlation, incorrect assumptions as to lags and over determination can each and all destroy a model’s usefulness.
This is particularly true when you do not know how one variable interacts with another and whether such interaction is mediated by a third which is not actually present in the model. In many cases you might start with a general hypothesis - “Incumbents are more likely to be re-elected when the economy is robust.” - and then have to qualify this with such things as “Unless there is an unpopular war on, unless the Prime Minister is proven corrupt, unless the party has been in power for more than 12 years.” Each qualification reduces the predictive capacity of the model which may, in fact, give you a great deal of information your original hypothesis lacked but which means you are less able to generalize or make causal claims.
Written by jay on February 5th, 2007 with
4 comments.
Read more articles on "Global Warming" and Canadian Politics.
- [+] Digg: Feature this article
- [+] Del.icio.us: Bookmark this article
- [+] Furl: Bookmark this article
doug newton
lrC
#1. February 5th, 2007, at 9:30 PM.
Do these work now?
Not so much a bore as I have no idea why this is the Gotcha flavour of the moment. Likely as it is such an unimportant topic on an immediate basis. All parties run to the environment as it can mean anything. All amateurs critique general theory as it is punching at shadows.
Now if the blogging libertarian democratic sorts (confess - something of an oxymoron) were all over countries we trade with who crush liberty and democracy, that would be something worth following.