Path: utzoo!attcan!uunet!husc6!bloom-beacon!gatech!uflorida!novavax!proxftl!bill From: bill@proxftl.UUCP (T. William Wells) Newsgroups: comp.ai Subject: Fuzzy systems theory was (Re: Alternative to Probability) Summary: fuzzy systems theory is all wet Keywords: fuzzy, logic, realism Message-ID: <183@proxftl.UUCP> Date: 17 May 88 19:25:39 GMT References: <4134@super.upenn.edu> <3200014@uiucdcsm> <1484@pt.cs.cmu.edu> <1196@bingvaxu.cc.binghamton.edu> Organization: Proximity Technology, Ft. Lauderdale Lines: 83 In article <1196@bingvaxu.cc.binghamton.edu>, vu0112@bingvaxu.cc.binghamton.edu (Cliff Joslyn) writes: > Furthermore, such a theory exists: Fuzzy Systems Theory. From what I have read in the popular literature, I think that fuzzy system theory cannot work. I wrote a letter to Byte recently in response to an article on fuzzy logic, and so, rather than thinking up another way to say the same thing, I am going to post it here (with some editing). === 4/5/88 Dear Editor: The article "When Facts Get Fuzzy", BYTE, April 1988, p. 285, contains fuzziness of its own: philosophical fuzziness. Take his example of "partly cloudy", which he asserts is ambiguous. As he presents it, it is, but the ambiguity is caused by the fact that "partly cloudy" is not an attribute of the sky but rather of some person's perception of it. A philosophically accurate (at least to a Realist) statement is: "partly cloudy" is the form in which some particular person perceives the sky when the sky contains clouds in certain configurations. People can say "partly cloudy" and be understood by others because the configurations which people see as "partly cloudy" are usually the same. Consider the statement "I see that the sky is partly cloudy." There is no ambiguity there; either I do see this or I do not and the statement is either true or false as a consequence. It might be the case that sometimes, given the same sky, I would report it as partly cloudy or not and one could measure a probability, but that probability is an attribute derived from myself, not from the sky alone. Consider the statement "The sky is partly cloudy." Since this statement omits WHO sees it as partly cloudy, it is ambiguous. One could take a group of people, ask each one whether he sees the sky as partly cloudy, and compute a probability; but that probability is an attribute of groups of people, not of the sky alone. The only way to fix the ambiguity, without making the error of attributing "partly cloudy" to the sky, is to make use of measurements of the density and distribution of clouds in the sky; these being attributes of the sky and independent of who is looking at it. After having done this one can then give a rigorous definition of "partly cloudy" and any statement about the cloudiness of the sky can be validated by using this definition and without reference to who is making it. If, on the other hand, you still would like to be able to say that the sky is X% cloudy and are unwilling to use specified measurements of the sky to do so, how would you proceed? The only way you could do so is by polling people, but of course, as I said, your statistic becomes one about groups of people. The same kind of reasoning applies to the idea of "young". If one specifies WHO thinks that someone is young, the ambiguity goes away; and if one does not, all you can get is a statistic which indicates what fraction of some group of people think that a person is young. Because fuzzy logic is based on a fallacy (the assignment of set membership based on a relationship, while disregarding the thing(s) being related to), any conclusion derived by the use of fuzzy logic is invalid. Cases where it appears to work are sheer dumb luck. I could go on -- but what would be the point? Although the article contains several more philosophical errors, they are irrelevant since the basic premise of the article is flawed. Beware bad philosophy! === So, is my perception of fuzzy systems theory an artifact of my sources, or does it really rest on the philosophic confusion of its proponents?