The Main Article and Home of Hubris
The Warning of von Neumann's Elephant
This web site dissects one really good example of how bad modeling works. As you see how it works, bear in mind that the lead author of the example was the Chief Scientist of a division of NOAA and others of its authors had international reputations.
This is a graph of a quantity Z with respect to a quantity N. As you can see without arithmetic, the correlation of Z to N is about 0. Yet, it has been strongly believed among a grouping of imminent scientists that Z increases with N. Advice has been given to governments based upon this belief, and turned into government action. You might ask: is it possible for an entire field of science, albeit a small field, to run, herd-like, into a blind alley? And to take part of the government of the United States with it? If you have enough sense of mathematics to understand this graph, you will be able to that the herd failed to see the trap. And their first error was in not including that graph in their article.
We will see in the main example-article, taken from science of the Gulf of Mexico, that "successful" multi-parameter curve-fitting swelled an already exaggerated certainty in a plausible but not proven theory.
In addition to the clear error of von Neumann's Elephant, the acclaimed article has other errors, detailed in Hypoxia or Hubris, that seem to have gone unnoticed by the "peers":
* mere curve-fitting of past events was mistaken for confirmational "prediction,"
The model, Scavia 2003, dissected in this web site, ought never have been taken seriously because is it clearly too simple for the phenomena it represented. At least one of the authors seems to have been pleasantly astonished that such a simple model could "predict" the existing data so successfully.
But, in its simplicity, the model implies impossible physical conditions. Had anyone thought carefully about its equations, this implication might have warned him to fix or to abandon the model. Rather, in the error that von Neumann had warned us against, the model's parameters were selected so that the model "successfully" matched some data. This matching of old data was called "prediction." The model and its theory were then assumed to have been confirmed well enough to be used in advising government.
The model implies that the concentration of oxygen is sometimes negative. Why would this not disprove the model?
The model implies that, for several numbers Y, Y is the same as Y - 1.
You can study this implication in the larger article Hypoxia or Hubris. For very practical reasons, no real science allows its theories to imply falsehoods.
As hinted at in the graph above, even before the computer-model existed, a proper graphing of its raw data suggested, had anyone un-blinded by reasoned expectation looked at it, that the theory is false.
The example dissected in this website, though complex in some ignorable details, might be the simplest of examples of widely acclaimed, peer-accepted bad science. Thus, it might provide a good learning experience and warning for thinking people - even those who have no particular interest in "dead zones." The errors might be trivial compared to some similar errors in other environmental sciences. But, they are accessible and suggest, clearly I think, that profound errors are being made routinely in modern science.
Being blinded by Belief-Based Hubris is a human condition. Scientists, too, are human. Some of our Beliefs are made particularly plausible by reasoning from empirical roots - but history shows that objectivity and rationality are not perfect defenses against powerfull, mind-melding social instincts.
The multiply-flawed model shows signs of being the product of a peer-clique that is blinded by a shared certainty. It
The societal success of the model, Scavia 2003, and of its theory of a single springtime cause of hypoxia in the Gulf of Mexico, remind us that people often move in mutually confirming "herds." I believe that participating in "herds" and "cliques" is instinctive in all of us, and, therefore, forgivable.
Study Hypoxia or Hubris? and find out.
Withering rebuttals are welcome: