biep@cs.vu.nl (J. A. "Biep" Durieux) (12/15/87)
I have lost the article by Robin Faichney (sp?) to which this one is a response. A list is more inclusive as it gets longer; a set of requirements is more exclusive. I never required a methodology to be optimal, so I do not require rules of every sort from my list to be included. I only said that >I don't think I can give a rigorous definition of "methodology", neither of >the discipline, nor of the product. Nevertheless I will give a try, perhaps >there are any takers.. > >A methodology is a set of rules that (in the ideal case) determine the >conduct of a scientist following it in the face of the facts and currect >hypotheses and theories of the (scientific) discipline the methodology >(and the scientist) is about. So merging this with the OED definition, this comes down to the statement that a method is a rule or a set of rules which determine conduct. (I am currently unhappy with the exclusively science-oriented viewpoint: as I have said elsewhere, I think methodologies can and do exist elsewhere.) Also note the "(in the ideal case)": I allow for some sloppyness in stating rules ("guidelines"), but I consider such cases non-ideal. My article continued.. >That for the abstract point of view. Bob Myers asked me to describe a >methodology for some actual science. I do not have the knowledge to do >so (I do not know when to resort to experiment, etc), but I will try >to describe the form of some of the rules of conduct that a methodology >(for many sciences) will have: >There are rules, stating: >- how to state facts >- how to order facts, givens . . . >- how and when to merge fields (e.g. some day chemistry might turn into > physics) So here I am summing up which rules and rule-sets (methods) scientists, in my opinion, actually use. Therefore, the list given should be seen as a list of examples of what sorts of rules I in fact do "allow". So the longer the list, the more inclusive my "definition". On the other hand, I think I am willing to hold, that any *realistic* *scientific* methodology *has to* contain at least a lot of these (unless some completely different new methods are invented): any methodology not containing them will, if followed, yield poor research. [No, I know a possible exception: a methodology that only says: "In whatever circumstance, do what your teacher tells you to" might yield good results if the teacher is good and willing to cooperate. I would consider that singleton-ruleset a valid methodology, and if it yielded good results a good one.] >Undoubtedly there will be many more aspects to a methodology, but I hope >this makes somewhat clear what I mean by the word. If people either want >to extend this list, or to fill in some "real" rules, by all means do so, >It will probably be very useful for my PhD project! (Cooperative planning >and problem solving, including learning, theory formation, etc.) My opinion is, that the more detailed one allows ones methodology to be, the better results it may yield. Breadth-first forward search will find the answer if there is one (and the branching factor is enumerable), and so a quite valid methodology, but clearly not a very good one. My hopes are to get a better one by allowing it to be "smarter". Perhaps I don't succeed, but for me that wouldn't mean a disproof of my ideas. >One might wish to add "practical methodology": >- how, when and where to apply for money >- how to decide which colleagues to trust (i.e. to accept their theories, etc. > without checking them) >- how often and when to rest >- how much, and in which direction, to specialise -- Biep. (biep@cs.vu.nl via mcvax) My F-key has autorepeat