I will focus my considerations on a point that, although it is of crucial importance for the thesis of the author, has not been made so explicit in the paper. The question is about a systematic ambiguity of the meaning of "meaning," and can be expressed as follows: does language denote things or thoughts? This is the basic question about meaning, and unfortunately it is rarely recognized; as a consequence, several not well-founded contrapositions arise due to the fact that the two semantic solutions are very different in regard to their consequences. Simon faces the problem on p. 22. In Simon's words: "Are there human absolutes, unchangeable from one culture to another [. . . .]?" This is the central point. Are human psychological schemata the same for all human beings? The question arises of the possibility of a full translation from one language to another. On this topic, one can see that Simon chooses the second solution: words denote thoughts, which are the same for all human beings; and this fact guarantees the possibility of communicating, and even of reaching a full translation from a language to another. This semantic solution, although not the most common, is strongly rooted in Western philosophy. Let us take a look at Aristotle's Peri Hermeneias:
Now spoken sounds are symbols of affections in the soul, and written marks symbols of spoken sounds. And just as written marks are not the same for all men, neither are spoken sounds. But what these are in the first place signs of-affections of the soul-are the same for all; and what these affections are likenesses of-actual things-are also the same [16a3].
Here the main theses are: A) signs, which are different from a language to another, are symbols of thoughts; thoughts are the same for every person. And B) thoughts represent things, which in turn are the same for every person.
I will not discuss here A); I limit myself to notice that A.1) such an assumption makes possible psychology as a science: the science ontically committed to the domain of possible mental structures. And that A.2) this invalidates the Whorfian hypothesis. With regard to B), this is the rub. The main dilemma about semantics lies here. Following Aristotle, thoughts represent things. As a consequence, you can have a well founded thought, and no need to express it. The alternative point of view started with Stoicism, and was developed by Nominalists (apart from the quasi-mythic Roscellinus, I have in mind Ockham, Hobbes, and all the greatest thinkers of 17th century): against B), they claim that language is a bridge between thought and world. As a consequence, you cannot have meaningful thought without symbols (see: Leibniz's Dialogus or Wittgenstein's Tractatus).
Under Aristotle's hypothesis, that is, starting from a conceptualistic point of view, the denotatum of a name is a mental representation, or a mental schema. In order to join language and things, no semantic solution can avoid, as a consequence, to cross thought. On the opposite side, from a nominalistic point of view, you can put thought inside a dimension preceding the semantic relation, which joins directly language and world. Thought, being prior to the semantic relation, can direct its attention, and can so select its object (by "intentionality"), and only later on an expression will fill the intention. It is only on the basis of this new semantic foundation that you can have a Kundgabetheorie der Sprache, as opposite to an Ausdruckstheorie.
Let me conclude. Simon's thesis seems to me to agree totally with Aristotle's semantic solution; on the contrary, traditional AI paradigms remain, in my opinion, with few exceptions (see, for instance, Schank's conceptual dependency theory) on the other side: we cannot avoid, while creating intelligence, the symbolic medium. Going on under this stimulating, very old and very up-to-date suggestion, can we develop (and test) a new line of AI applications, based on new semantics assumptions? On the one hand, as we must develop programs running on a computer, obviously we cannot avoid a linguistic medium; but, on the other, we could distinguish between a lower level, symbolic structure, and a higher-level simulation of a conceptual, non-linguistic, directional and intentional deeper knowledge. The model I have in mind might be a common field of work for both cognitivism and connectionism. Put language on the one side, and things on the opposite side. Thought is the medium; it is what language actually expresses; it is the denotatum of Language. On the other hand, Thought represents the real World:
L --> T --> W
Well, the first arrow represents cognitivism: the way in which language refers to thought cannot avoid being conscious, that is, to be organized by explicit, logical, formal rules (that is a condition for being expressed or communicated); with regards to the second arrow, what is the natural manner in which thought can represent things, having no more logic and language? At the hardware level, we find here physiological, neural configurations; at a software one, here we have ta pathemata, the mental configurations, or the complex structures we passively assume in facing things. Thus, the link between language and world becomes much more sophisticated: on the left side, one could represent knowledge by means of formal order logic, or of other techniques; on the right one, one could use neural networks, or other sophisticated structures. In this sense, all the applications which traditionally developed only one side of the question have forced the other to be in charge of the tasks left untouched.
Even though it is clear that Simon's paper is not focused on the point I have discussed here, I suppose he could agree with me that it is of crucial importance not only for the semantics of literary criticism, but, perhaps even more, for the future development of AI itself. Under this suggestion, a new approach could be built in between what I called the left and the right sides of the semantic problem; that is, new mixed techniques of knowledge representation could arise, assuming that two different processes would collaborate: the one having the goal to use language for representing thoughts, and the other, assuming thoughts for representing things.
1. I have no space here to qualify why I disagree with the English translation of "affections of the soul" for the Greek "pathemata tes psyches." In short, I propose "structure" for "pathema" and "mind" for "psyche."
2. According to the famous Whorfian hypothesis, language affects cognition (see Whorf, 1956 and Schank, 1975: 9). For modern arguments against the Whorfian Hypothesis see Heider and Oliver (1972), and Glocksberg and Danks (1975).