Recipes for the Simple-Minded: a Constructive Account of Original Intentionality
To mentally represent the way things are, a subject must be capable of getting things right or wrong "by its own lights," or with respect to standards acknowledged by that very subject. Although maps, calculators, and maybe even frogs possess a derived sort of intentionality that is parasitic upon the standard-setting capacities of others (or Mother Nature), they lack the original sort of intentionality that should interest those studying bonafide mental capacity. Recently, Dennett has denied that anything could exhibit this more interesting, original form of intentionality, arguing in effect that nothing, not even persons, can be understood to be beholden to the way things are, without ultimately appealing (as we do with frogs and slot machines) to the background of purposes for which a subject has been selected or designed. Dennett thus raises an intelligible challenge to intentional realists, one that has not been adequately addressed by most naturalistic theories of mental representation: how can something be correct or mistaken about the way things are, where the standards in play are intelligible as such, without appealing to the purposes for which that thing has been designed or selected?
In my dissertation, I respond directly and constructively to this challenge by profiling in suitably naturalistic vocabulary how critters could comport themselves, so that they can be correct or mistaken with respect to standards funded out of their own responsive dispositions. As I see it, intentional capacities arise as critters acquire resources to cope flexibly with their environments, resources that can be described in suitably unmysterious terms. In particular, I describe how a critter’s procedure (or responsive dispositions) could be governed in part by the activation of certain cognitive structures I call ‘expectations.’ By modifying these structures in the wake of what I term their ‘violation,’ expectation-mongering critters are able to exhibit an educable capacity sufficient to meet Dennett’s challenge. However, to sustain that claim, I must show how critters with such a capacity are beholden to the way things are in a way that doesn’t ultimately appeal to their selective purposes, something that most other attempts to ground special intentional capacities in educability (e.g., Dretske's and Bennett's) fail to do. To this end, by explaining how we can understand such critters as engaging in a kind of rational goal-directed activity that is intelligible quite apart from their designed purposes, I show how expectation-mongerers are susceptible to two distinct types of error, corresponding to two ways in which expectations can fail to pick out regularities in a critter’s environment.
Not only does my description of expectation-based educability meet Dennett's challenge head-on, this account also has the resources to describe progressively more sophisticated types of adjustments to environmental contingencies. It scales up much better than the accounts favored by most learning theorists. By adding more structure to this account, I show how critters with more sophisticated educable capacities can make progressively more subtle kinds of errors. I can therefore describe several distinct kinds of intentional activity, corresponding to different ways in which a critter can be more or less "on" to its environment, including primitive ways in which critters in a community can become beholden to one another. Although I don't pretend to render intelligible our most sophisticated intentional capacities - those inferential and linguistic capacities in virtue of which we are able to reason with one another about how the world is - my account nevertheless shows that there are interesting pre-linguistic boundaries separating the relative intentional capacities of critters, boundaries that are liable to be overlooked by those (like Davidson) who insist that thought or reason is in some sense limited to linguistic beings. This project, since it articulates the boundaries around the pre-linguistic intentional capacities from which our full-blown conceptual capacities plausibly evolved, is of evident value for the study of animal cognition, and should prove useful for the task of setting reasonable targets for AI research.