Some interesting points came up during the discussion after my talk on Tuesday (the handout and audio for which should be up soon on the E-Intentionality page/RSS feed).

In the talk, I had made a distinction between three kinds of Intentional Systems Theory:

- Modest: Determines whether something is a true believer (TB)
- Ambitious: Determines not only whether something is a TB, but if so, also determines what propositional attitudes they have
- Moderate: Determines whether something is a TB, and in some cases, also determines what propositional attitudes they have

I then claimed that although non-transitivity seems possible in the case of modest (or moderate) IS theory, it isn’t possible in the case of Ambitious IST. Here’s the reasoning behind my claim. Consider a purported case of non-transitivity: my best interpretation of x is that x is an intentional system, x’s best interpretation of y is that y is an intentional system, but my best interpretation of y is not that it is an intentional system. (Actually, the are problems for Ambitious IST even in some cases where my best interpretation of y is that y is an intentional system; see below.) In such as case, Ambitious IST takes y to have the propositional attitudes ascribed to y in x’s best interpretation of y. Therefore, my interpretation of x will have to involve not only ascribing to x certain propositions P1 that x actually has, but also ascribing to x the capacity to entertain propositions P2 (propositions that x might not be entertaining at present) that constitute x’s best interpretation of y. For x’s best interpretation of y to be that y is an intentional system, believing/desiring the propositions in P2 must constitute the ascription of a set of propositions P3 to y. The kind of non-transitivity I had in mind was one where the concepts used in P3 are not possessed by me, thus preventing me from interpreting y in the way that is x’s best interpretation. Could it really be that I possess the concepts used in P2, but not those in P3, even though believing the propositions in P2 constitutes ascribing the propositions in P3?

I think now that there are several problems with this argument.

Start with the final question: the answer now seems to me to be “yes”. I can ascribe a concept C to someone even if I don’t possess the concept C; this is because possessing a concept is not required for referring to it. So x can ascribe to y concepts that x doesn’t possess, and thus I need not ascribe to x any concepts that I do not possess; I need only ascribe to x those concepts necessary for referring to y’s concepts (in x’s best interpretation of y). So yes, this kind of intransitivity is possible. Or at least it has not been shown to be impossible.

Second, other kinds of intransitivity, not argued against, may be possible. That is, despite the fact that x’s best interpretation of y is that y is intentional, it may be that my best interpretation of y does not take y to be an intentional system at all, even though I possess all the concepts x possesses (and even all the concepts x’s best interpretation of y attributes to y).

Similar considerations suggest a slight variant of the argument, which I may have sometimes shifted to during the talk. The variant considers a different kind of non-transitivity, where my best interpretation of x is that x has propositions P1 and the capacity to entertain propositions P2 that constitute x’s best interpretation of y, x’s best interpretation of y is that y has propositions P3, but my best interpretation of y ascribes to y propositions P4 distinct from P3. The argument would be that this is a contradiction, or at least an undesirable indeterminacy, since y’s propositional attitudes are supposed to be both those in P3 and those in (distinct, incompatible) P4.

One might think that Ambitious IST can be patched up, by adding some kind of rule to resolve this indeterminacy. For example, one could say that my propositional attitude ascriptions, if such are present in my best interpretation of system y, have priority. Or, to handle indeterminacies between two interpreters that are not me, we could say that the ascriptions of the interpreter that is closer to me in the chain of “best interpretable as intentional” relations have priority. But what if the interpreters are equi-distant? And if we are going to merely defer to me as to what is real, then why bother tying TB-status to anyone’s capacity to interpret other than mine? Why not just say that something is a TB (or in the class of interpreters that define TBs) if I can interpret as such, and leave it at that?

Another kind of patch for Ambitious IST is to reject considerations of conflicting best interpretations as practically impossible (cf Dennett’s remarks about the “cryptographer’s constraint”).

One general point that should be kept in mind when considering these proposals is that conflicts between two interpreters’ best interpretations should not arise because of mistakes or differing sets of information. The idea is that all interpreters are “functioning properly” and have access to the same set of data. Admittedly, there may be problems in the notion of interpreters with distinct sets of concepts somehow being given the same data, relying as the notion does on a sharp observation/theory distinction.

Also, one might wonder if anything other than Ambitious IST makes sense. Can one really establish that something is best interpretable as intentional without having that interpretation to hand?

Comments/questions welcome.

Ron