kck@g.gp.cs.cmu.edu (Karl Kluge) (01/10/90)
It occurs to me that Searle's argument (in particular his 3rd axiom, "Syntax by itself is neither constructive of nor sufficient for semantics") depends on the reader accepting as plausible and natural a rather bizarre claim. Specifically, in defense of this axiom on page 31 he says, "But now imagine that as I am sitting in the Chinese room shuffling the Chinese symbols, I get bored with just shuffling the -- to me -- meaningless symbols. So, suppose that I decide to interpret the symbols as standing for moves in a chess game. Which semantics is the sysytem giving off now? Is it giving off a Chinese semantics or a chess semantics, or both simulaneously?" Well, this is pretty strange if one assumes that once Searle decides to start interpreting the symbol shuffling as a chess game, the Chinese speaker outside the room continues to see what looks like a coherent conversation in Chinese. Are we to suppose that you can impose an interpretation on the symbols such that they form both a coherent conversation in Chinese *and* a legal and coherent chess game? Suppose we had a Chinese speaker who didn't understand English inside the room, shuffling English text and producing outputs that mimic those of a native speaker of English. Suppose you were standing outside the box, having what to you looked like a perfectly sensible conversation in English, and up walks Searle and insists that what's really going on is a chess game. Would that make sense to you? Let's picture a more simple situation, the addition-mod-16 room. You pass in triples of characters from the set (0-9, a-f), and out come triples of characters from the same set. Inside the room is an individual with the following rules: If the input is not of the form (x,y,z), where x, y, and z are characters from the set (0,1,2,3,4,5,6,7,8,9,a,b,c,d,e,f), then hit the big red button and do nothing until another input comes in. If the input string has the correct form, then match x with a column heading in the following table, and y with a row heading in the table. Trace along to the intersection of the row and column. If the character there doesn't match z in the input string, press the big red button; otherwise output the string (y, z, w), where w is the character in the table in the row and column specified by y and z. [Addition mod 16 table omitted for brevity's sake] Suppose you are standing outside the room. If you pass in a triple which specifies an incorrect sum mod 16 (such as "6, 5, 7"), or an improperly formed string (such as "foobar"), you hear a loud buzzer. If you pass in a triple which does specify a correct sum mod 16 (such as "6, 5, b"), out comes another triple specifying a correct sum mod 16 (in this case, "5, b, 0"). Here you are, having a coherent (if somewhat boring) conversation about addition mod 16, when along comes Searle. He insists that the triples (a,b,c) are to be interpreted as follows: a specifies a chess piece (of which there are 16 on a side), b and c taken mod 8 specify the row and column on a chess board that the specified piece is to be moved to. What's going to happen? Can Searle come along and just impose any semantics he likes on the symbols this way? Well, sometimes Searle will pass in a perfectly legal move in his interpretation, and all the room will do is buzz at him. He can pass in illegal moves, and the room will cheerfully respond with a move. The moves the room responds with will sometimes not be legal moves, and in general those which are legal moves won't make much sense. Is it sensible to say that the triples have the semantics Searle claims? (Remember, we're not talking about "semantics" in the sense that someone who does formal logic does -- in fact, if Searle is correct, we *can't* be talking about the same kind of "semantics".)
kp@uts.amdahl.com (Ken Presting) (01/10/90)
In article <7502@pt.cs.cmu.edu> kck@g.gp.cs.cmu.edu (Karl Kluge) writes: >It occurs to me that Searle's argument (in particular his 3rd axiom, "Syntax >by itself is neither constructive of nor sufficient for semantics") depends >on the reader accepting as plausible and natural a rather bizarre claim. >Specifically, in defense of this axiom on page 31 he says, "But now imagine >that as I am sitting in the Chinese room shuffling the Chinese symbols, I >get bored with just shuffling the -- to me -- meaningless symbols. So, >suppose that I decide to interpret the symbols as standing for moves in a >chess game. Which semantics is the sysytem giving off now? Is it giving off >a Chinese semantics or a chess semantics, or both simulaneously?" > >Well, this is pretty strange if one assumes that once Searle decides to >start interpreting the symbol shuffling as a chess game, the Chinese speaker >outside the room continues to see what looks like a coherent conversation in >Chinese. Are we to suppose that you can impose an interpretation on the >symbols such that they form both a coherent conversation in Chinese *and* a >legal and coherent chess game? ... (cogent example deleted) > ... (Remember, we're not talking about "semantics" in >the sense that someone who does formal logic does - in fact, if Searle is >correct, we *can't* be talking about the same kind of "semantics".) I think Karl has a crucial point here. Searle's claim that "syntax is neither constructive of nor sufficient for semantics" is true when applied even to formal semantics, but does not have the strong implication Searle needs. What's true is that no syntax determines a unique semantics, even when axioms and inference rules are specified. That's how we get non- standard models of the real numbers, and other non-standard interpretations of mathematical theories. But Karl is right to point out that there are limits to the interpretation of a formal grammar.