My post about how to distinguish real science from fake science generated some interesting comments, especially over my assertion that we have not as yet been able to come up with an unambiguous way to distinguish science from non-science, and it prompted me to post a follow-up to try and clarify it.
For those interested in this demarcation problem, a good starting point is Larry Laudan’s 1983 paper The Demise of the Demarcation Problem. It is not available online (as far as I know) but it has been reproduced numerous times in various books and is not hard to find. The book But is it science? (edited by Michael Ruse) has papers, including Laudan’s, that argue both sides of this controversy. This post summarizes Laudan’s argument and addresses some of the criticisms of it.
Laudan points out that a satisfactory demarcation criterion must haveboth necessary and sufficient conditions. When people think that it should not be too hard to find ways to distinguish science from non-science, they usually do so on the basis of specifying one or the other, not both.
Why are both necessary? If you specify only a set of necessary conditions, then if something does not satisfy all of them, then you can say definitely that it is not science. But if something satisfies them, all you can say is that it could be science but not that it definitely is. For example, take the necessary condition that a scientific theory must be falsifiable. The theory that the moon is made of green cheese is falsifiable so it meets the necessary condition. But is it science?
Conversely, if one specifies a set of sufficient conditions, then if something satisfies all of them, then one can say definitely that it is science. But if it fails to satisfy some of them, then you cannot definitely say that it is not science. It may still be science. Again, suppose we create a set of sufficiency conditions of which one is the requirement that a theory must be falsifiable. Currently string is not currently falsifiable since it does not make specific predictions that can be tested. So that means that we cannot definitely say that it is science though it may well be.
The requirement that one must have both necessary and sufficient conditions for suitable demarcation criteria is generally true. But there is also an auxiliary condition at play. When we are creating demarcation criteria for something, we are chiefly interested in a class of objects (or theories) for which we already have some idea of what belongs and what does not. If we create demarcation criteria that fail to include members that should obviously belong (based on our intuitive sense) or includes those that obviously should be excluded, those criteria would not be suitable either. Demarcation criteria that ended up classifying general relativity as not science and astrology as science would be looked at askance.
As a trivial example to illustrate these points, suppose that I want to be able to distinguish between chairs and non-chairs. I could specify a necessary condition that a chair must have four legs. If something has three legs, then it is definitely not a chair, according to my criterion. But if it has four legs, all I can say is that it may be a chair or it may not. It may be a dog.
Similarly, suppose that I specify a whole set of detailed conditions for a chair (it must have four legs, have a horizontal seat and an upright back, and be at a height suitable for sitting on) and say that any item that meets all those conditions is sufficient to certify it to be a chair. Undoubtedly many traditional chairs would meet those criteria. But what about a chair made by incompetent carpenter in which the seat is not horizontal or the back is not upright? It does not meet the sufficiency criteria and so we cannot certify it to be a chair. But it may well be a chair. Many modern chairs with their unusual geometries may also be excluded. In fact, the more rigorous one makes the sufficiency criteria, the more likely one is to have things that one thinks should belong not make the cut and be left in limbo.
And what about a sofa or a bed? They would meet all the sufficiency criteria to get the certification of being a chair though neither is what we normally consider to be a chair. So this fails the auxiliary test too.
That is why the necessary and sufficient and auxiliary conditions are required for satisfactory demarcation criteria. With chairs it may be possible to construct such conditions but Laudan thinks that this problem is insoluble for the science/non-science split. Others have challenged his view but not (in my view) conclusively. But his essay is definitely an excellent starting point for understanding the issue.