2 ms·
The general principle is fine. The principle (stated more precisely) is, "if you have an accurate mathematical model, you can do engineering." In the case of c
by yummyfajitas 17y ago
The general principle is fine. The principle (stated more precisely) is, "if you have an accurate mathematical model, you can do engineering."
In the case of cancer, we don't have an accurate model. All we managed to do is use a large statistical sample to average out over the non-radiation variables.
In the case of climate modeling, we have a sample size of 1. All our conclusions come from a family of models which are claimed to be very accurate. According to the IPCC, our models are accurate enough to make the claim (with 95% confidence) that CO2 and not other unforeseen factors is the cause of the current global warming trend.
If our model is that good, what prevents us from using it for engineering?
And if the model misses all sorts of important facts, how can we draw conclusions like "AGW is true" from it?
- philwelch 17y agoThe sample size is >1: global warming through high CO2 has been observed on the planet Venus. Our model for radiation poisoning is pretty good, but mechanistically there's no way in that system to easily undo the damage of radiation. Also, it's disingenuous to say that the sample size for climate modeling is 1. We have multiple samples, from multiple methodologies, of the global temperature and the global atmospheric CO2 throughout time. We also have multiple samples of experimental data confirming and quantifying CO2's contribution to the greenhouse effect. Your argument is kind of like saying, "we know bombs work, we know how bombs work, and we know that all of these bombs we've been dropping on Berlin are positively correlated with the increasing destruction of the city, but there's insufficient evidence that the problem will get any better if we stop dropping bombs on Berlin. So let's build a gigantic dome over Berlin made out of nanotubes." To do engineering on a system, you have to know nearly everything about how all of it works. To diagnose a single problem you can narrow down your knowledge a lot more and have much higher certainty. So even your restated principle is wrong. We know that not irradiating yourself will lower your risk of cancer. We don't know how to undo the damage caused by radiation. Again, your argument is kind of like saying, "if we're not sure how to fix radiation damage to tissue, we have no proof that radiation has negative health effects". Furthermore, your decision theory is wrong too. We have to be 99% confident that geoengineering won't cause any bad, unforeseen side effects before we even do it. But we are even 50% confident that something we are doing (emitting CO2) is having a potentially catastrophic effect, we have to act on it.
- yummyfajitas 17y agoThe sample size is >1: global warming through high CO2 has been observed on the planet Venus. 2 is not large enough for statistical averaging, though it can improve our confidence in our models. Our model for radiation poisoning is pretty good, but mechanistically there's no way in that system to easily undo the damage of radiation. If climate also lacks a path from state A to state B, the models will reflect it. That's a question answerable by a few supercomputer runs (if the models are correct). Also, it's disingenuous to say that the sample size for climate modeling is 1. We have multiple samples, from multiple methodologies, of the global temperature and the global atmospheric CO2 throughout time. We also have multiple samples of experimental data confirming and quantifying CO2's contribution to the greenhouse effect. No, we have precisely 1 sample: earth. We have another sample (Venus) in a vastly different region of parameter space. Multiple measurements of a single sample is not a substitute for multiple samples. To do engineering on a system, you have to know nearly everything about how all of it works. To diagnose a single problem you can narrow down your knowledge a lot more and have much higher certainty. We cannot conclude that current global warming trends are anthropogenic without excluding all other reasonable factors. Thus, the models must have control over those other factors. If we had a larger sample (multiple earths, or multiple patients in your analogy), we wouldn't need this degree of control. But we don't have such a sample. Incidentally, your analogy of dropping bombs on Berlin illustrates perfectly the use of an accurate model. We have accurate models of bombs and their effect on structures and vehicles. The accurate models make engineering (structural, in this case) possible, including the design of bomb-resistant buildings. The problem is now reduced to a simple question: which costs less, reducing the rate of bombing or building bomb-resistant buildings?
- philwelch 17y ago"No, we have precisely 1 sample: earth. We have another sample (Venus) in a vastly different region of parameter space. Multiple measurements of a single sample is not a substitute for multiple samples." The idea of large sample sizes is to make intelligent inferences about a large and diverse population. We have one planet--anything that tells us about how this planet, in particular, operates is useful, and "low sample size" is a bullshit criticism. So is our chemical understanding of how light interacts with certain gas mixes, particularly CO2, in the realm of heat retention. "We have accurate models of bombs and their effect on structures and vehicles." We have accurate models of CO2 and its effect on heat retention as well. We don't have to rule out the idea that fairies are sneaking into Berlin and destroying buildings with pixie dust when we know how many bombs we've dropped on the place, can survey the amount of destruction, and can make intelligent conclusions about whether the number of bombs dropped matches the amount of destruction. "The accurate models make engineering (structural, in this case) possible, including the design of bomb-resistant buildings." Yes, and your proposed solution is to relocate the entire population of Berlin into underground bunkers, whereas my proposed solution is to stop trying to conquer Europe, since we shouldn't have done it in the first place and now it's causing terrible consequences for us.