I found Gigerenzer’s argument for the ecological rationality of heuristics and his argument against content-blind norms to be compelling. Specifically, I’d like to share some of my thoughts about content-blind norms.
Intuitively, it seems to make sense that rationally parsing meaning should be a content-blind process. After all, if it weren’t, wouldn’t content-dependent interpretation already require a preconception of the content’s meaning? If so, it seems this would amount to infinite regression, or a recursively defined meaning-generating function that accepts meaning of the sentence as input.
Let M be a function that maps from a sentence s to a meaning m (so we could say M(s)=m). Let D be the incomplete meaning-generating function from the above paragraph, where D maps a proposition (P) and meaning (m) to a (hopefully more complete) meaning. With this, we could write M(s) = D(P, M(s)). To clarify, proposition P corresponds to the content-blind syntactic structure of sentence s, while the meaning is the complementary context-sensitive component of sentence s.
Clearly, function D doesn’t do anything except call itself ad infinitum. It does, however, indicate what properties a correct semantic interpretation function C must possess: a base case for generating meaning that requires no precomputed meaning, and a means of reducing the extent of precomputed meaning required. That is, C(P, ε) must be defined, and C must satisfy M(s) = C(P, M(s’)), where s’ is a “less meaningful” sentence than s in some way.
The necessary existence of a base case suggests that at there must be some class of sentences whose meanings are entirely content-blind. I take this to mean that rationality is ultimately rooted in some fundamental set of processes that are entirely context-blind. That is not to say, of course, that rationality can be defined by content-blind norms—only members of this fundamental set can be defined by them.
I do not think this conclusion contradicts Gigerenzer’s view on viewing rationality in terms of content-blind norms. As Matheson puts it, “that cognitive virtue cannot be located entirely within the mind does not imply that it is located entirely outside the mind” (143). Supposing content-blind norms alone cannot be used to measure rationality, this question arises: can solely content-sensitive along with content-blind norms be used to measure rationality? For those two types of norms to suffice would require rationality to be purely self-blind. That is, for a given sentence, there would be one unique rational way to interpret it (ignoring situations of ambiguity). But is this the case? Each individual obviously interprets identical sentences in nonidentical ways. If rationality is self-blind, then, although no one may actually be 100% rational, we could conceive of someone who is 100% rational, the epitome of human rationality norms. Moreover, this would suggest that there could exist no two people who are fully rational yet with different minds.
I think it likely that rationality is also self-sensitive; that is, rationality needs to be evaluated with respect to the (ir)rational agent.