zarnuk@caen.engin.umich.edu (Paul Steven Mccarthy) (04/12/90)
>(Ken Presting) writes: >The [...] problem with explanations and understanding is that >we normally suppose that someone who provides an explanation is not just >exhibiting a skill which is *additional* to the expertise they display >in practice. We expect that when Kasparov explains chess, the ability to >explain is *part* of his ability to play. Oops, more of those durned assumptions. They seem to be roaming all over this newsgroup. >Paul's example of the program with the bolt-on explanation feature shows >that this expectation is simplistic. The conclusion to be drawn, I think, >is not that the capacity to explain one's choices is irrelevant, but >that the capacity to explain is not a sufficient condition for what we >call "understanding". >As far as I know, there have been no elegant solutions to the problem >of stating such a sufficient condition, either in the case of intentional >action or of understanding. I would take the issue one step farther. I think we are trying to pile too many different conotations under the single label of "understanding". So far we have seen (on this newsgroup at least): 1. Ability to perform successfully. (win playing chess) 2. Ability to explain the choices made during the performance. (Why did you win/lose?) 3. Ability to perform similar tasks. (8x9 chess-board) **4. The actual algorithms used to perform the task. (Some algorithms "understand", others do not). ** 5. Knowledge of the "strategy" of the game, and the ability to put that knowledge into practice. 6. Ability to reason about example games and formulate chess "strategy" from them. 7. Knowledge of the history of the game. (Who was the world champion in 1962?) ** (I _strongly_ disagree with this one, but it was presented here, and included it for completeness). I'm sure each of us could add many other abilities / constraints on the term to extend this definition to "capture the essence and subtlty" of the fuzzy notions that we carry in our heads. I'm just as sure that no two definitions would be exactly alike, and I have a strong feeling that no intersection or union of the definitions would satisfy everyone's intuitive notion of what "understanding" means. Item #4 above is a good example of why I think we would fail to generate a universally acceptable definition of "understanding". Why don't we just face up to the facts? The term, "understanding", is an over-simplification. Excuse me for my empiricist thinking, but let's stick to concepts that we can define unambiguously -- preferrably with measurable characteristics. I know this is contrary to a lot of the "new" thinking these days, but I think most of this "new" thinking is just "old" sloppy thinking. Those big, fuzzy terms. They're so *easy* to use, and they make you feel great too. You try rubbing them out, and soaking them out, and still... > [...] Just trimming the fuzz >on the concept of "understanding" is very tough. The "precise account" >I have in mind for "understanding" does *not* preserve the intuitive >connection between understanding an activity and being able to perform >it. I would be very surprised if *anyone* could provide a generally satisfactory definition for the term. (I apologize for putting you on the hot seat, but you did say...) >Fortunately, the Chinese room argument can be re-stated without any >appeal to the concept of "understanding". Isn't "understanding" the point of the Chinese Room? My point is that Searle et. al. are chasing fantasies. ---Paul... (Stick to the classics, you know, Gallileo, Gauss...)
kp@uts.amdahl.com (Ken Presting) (04/14/90)
In article <1990Apr12.165907.28054@caen.engin.umich.edu> zarnuk@caen.engin.umich.edu (Paul Steven Mccarthy) writes: >>(Ken Presting) writes: >>As far as I know, there have been no elegant solutions to the problem >>of stating such a sufficient condition, either in the case of intentional >>action or of understanding. > >I would take the issue one step farther. I think we are trying to pile >too many different conotations under the single label of "understanding". >So far we have seen (on this newsgroup at least): > > 1. Ability to perform ... > 2. Ability to explain ... > 3. Ability to perform similar tasks ... >**4. The actual algorithms used to perform the task. ... > 5. Knowledge of the "strategy" ... > 6. Ability to reason ... > 7. Knowledge of the history ... > >** (I _strongly_ disagree with this one, but it was presented here, and > included it for completeness). Kihong Park's suggestion that we drop the issue and attend to less confusing, but still challenging, examples of cognition should be mentioned also. >Why don't we just face up to the facts? The term, "understanding", >is an over-simplification. Excuse me for my empiricist thinking, but >let's stick to concepts that we can define unambiguously -- preferrably >with measurable characteristics. Paul has performed a valuable service by summarizing the points of view expressed so far. I think there is a useful comparison between our current discussion of "understanding" and foundational debates in other sciences. Here is an overworked but still handy example: The concept of "continuous function" went through a similar process. Nowadays we have several precise concepts all on the same theme - continuity at a point, continuity over an interval, uniform continuity, differentiability, infinite differentiability, and analyticity. *Any one* of these concepts would be an over-simplification. "Continuity at a point" is in some ways the most basic, but by no means do the rest of the concepts reduce to it. Uniform continuity, for example, is not a local property of a function, and the relation between diff'blty and analyticity is dependent on the space being mapped. (I see an interesting analogy between #4 and Stone's representation theorem, but I have a serious metaphor problem) I also see a pattern emerging <choke> from the various suggestions: Understanding is a relation between knowledge and abilities which is *sometimes* attributable on the basis of verbal behavior. >> [...] Just trimming the fuzz >>on the concept of "understanding" is very tough. The "precise account" >>I have in mind for "understanding" does *not* preserve the intuitive >>connection between understanding an activity and being able to perform >>it. > >I would be very surprised if *anyone* could provide a generally >satisfactory definition for the term. (I apologize for putting you >on the hot seat, but you did say...) I sure did, and there's no need to apologize! I claimed to have an account in mind because I have already swiped one from Quine! The whole point of the discussion of "radical translation" in _Word and Object_ (viewed in the context of the discussion here) is to give a definition of understanding a foreign language. As promised, having a radical translation ala Quine does not quite guarantee that one will be able to conduct a conversation in the foreign language (you may not be able to pronounce the clicks and pops). Quine's theory is not completely formalized, but it is careful enough to be called "rigorous", if the term is applied with some charity. Davidson's _Radical Interpretation_ is also intended to apply to language understanding, but I think it can be very naturally extended to apply to understanding any rule-governed activity. One of his basic ideas is "how does the speaker deal with his own mistakes". I propose that this concept should have the same role in a (future) family of formal concepts of understanding, as "continuity at a point" has in the concept of continuous function. (I promise to do my best to deal intelligently with my own mistakes, and I promise to make *plenty* of them. I think that a system which never makes mistakes *cannot* exhibit understanding. :-) >>Fortunately, the Chinese room argument can be re-stated without any >>appeal to the concept of "understanding". > >Isn't "understanding" the point of the Chinese Room? The relation between the concept of understanding and the Chinese Room is like a function which is everywhere continuous and nowhere differentiable :-). It's a counterexample, which is useful in foundational debates to show that a proposed definition is over-simplified. I will grant immediately that Searle has not presented it this way. We do not have any obligation to him to use his argument only for his purposes. If we can turn it to our own purposes, so much the better. The rulebooks of the CR, as Searle describes them, would not enable the operator to answer such obvious questions as "what color is the paper on which this question is written". This, I submit, is an interesting omission. Another observation: Searle, in his English-speaking persona, would not be able to learn Chinese semantics from his rulebooks, but he *could* learn Chinese syntax from them. If his rulebooks were enhanced so that he could answer practical questions, then he *could* learn the meaning of Chinese words from the books. An argument to this effect could be conducted entirely within Quine's theory of translation. (I should write this up) Isn't it better to beat Searle at his own game, and take all his marbles, and melt them down for CRT screens, than to cry "those marbles are sour?" >My point is that Searle et. al. are chasing fantasies. > >---Paul... (Stick to the classics, you know, Gallileo, Gauss...) We are *all* chasing fantasies - unless you can offer a proof of the the law of mathematical induction ... Ken Presting ("Counterexamples are the best examples")