Saving the Hypothesis August 22, 2013Posted by larry in economics.
An Inherent Complication in Assessing Tests of Scientific Hypotheses
The testing of scientific hypotheses is not as straightforward as it often looks and perhaps contributes to the fact that many politicians do not pay as much attention to the character and context of scientific evidence as they should.
Most of us are familiar with the probabilistic character of many scientific hypotheses and the ways in which these can affect the testing scenarios. But what is not as well-known is a strategy for “saving the hypothesis” that is independent of probabilistic considerations. This strategy makes the testing of scientific hypotheses more complicated than Karl Popper and others thought it was.
The methodology described briefly by the reviewer is that developed by Karl Popper, which is that scientists should concentrate on falsifying their theories rather than corroborating them. Many ordinary scientists use Popper’s falsification model as their guide. Effectively, what you do is list the initial conditions, and the lawful regularities, and in the case of the more formalized sciences, you deduce consequences from these, for instance a particular event – a procedure that has been discussed in physics and in other more formalized disciplines. If the event predicted by the theory under test does not occur, that is, is not observed, then Popper concludes that the theory has been decisively falsified. If the event does occur and has been observed, the theory has only been corroborated, that is, the likelihood that it is true has increased. For strictly logical reasons, no theory can ever be conclusively verified, that is, proved to be true, though Popper never explicitly points this out.
There is, however, a kind of get out of jail free card that can be used to “save the hypothesis”. This is known as the Duhem-Quine gambit. When you set up your theory testing scenario, in addition to listing all the initial conditions that apply and the regularities that are involved, there are a number of auxiliary hypotheses that usually go unstated with respect to the testing scenario. In physics, this is because experimental physicists often know what these possible contaminants of the testing process are and attempt to control for them. In less formalized sciences, such influences may be unknown or only informally considered. The Duhem-Quine gambit, when utilized appropriately, is a legitimate procedure, not an attempt to cheat the testing process.
Basically, if the result of the test of a given scientific hypothesis is not observed, for whatever reason, instead of directly falsifying the hypothesis or theory under test, the testers can select one or more of the auxiliary hypotheses as the culprit, such as the nature and operation of the equipment, any presumed biases on the part of the testers, sampling problems, the general environment in which the test takes place, and so forth. Selecting one or more of the auxiliary hypotheses that inevitably accompany any scientific testing situation will enable the scientist to “save the hypothesis”. Of course, one cannot continue to blame failure to obtain a result the theory says should have been observed but wasn’t on the auxiliary hypotheses. Continued failure to observe a result predicted by a theory must eventually lead to that theory (or a portion of it) being considered to be falsified.
Elliman’s point that more scientists should be involved in independent ways in policy decisions not subject to direct ministerial control becomes even more salient in light of the inevitable availability of the Duhem-Quine gambit, a central feature of any scientific testing scenario, which renders assessment of the relevance of scientific evidence for policy purposes even more problematic than it might otherwise be considered to be.