Dretske on Reducing Intentionality (Covariance)

?
  • Created by: temazcal
  • Created on: 26-05-18 18:05

Dretske's Teleological Account

1. Covariance

Dretske offers an information theoretic propsal. A device that carries information exhibits some degree of intentionality. Draws on Grice's notion of natural meaning. 

In essence: device S carries infromation about instantiations of property G iff S's being F is nomically correlated with instantiaons of G. 

If S would not be F unless property G were instantiated, then S's being F carries info about/indicates 'Gness'.

Spots on a human face, for example, carry information about a disease. A compass carries information about the location of the North Pole.

In all such cases, a property of a physical device nomically covaries with some physical property instantiated in its environment.

D1) The semantic content of a cognitive state M is a privileged part of its informational content, viz. that informational content of M which is nested in no other informational content of M.

D2) A cognitive state m of o has the proposition P as an informational content if the conditional probability that P, given that O is in M, is 1.

M's covariation with P's holding is not merely evidence that M has p as its informational content: it is constitutive. 

The Problem of Misrepresentation:

  • Follows from D2 that if P is the informational content of M, then P is true: so, by D1, is P is the semantic content of M, then P.
  • Solution: idealisation. Distinguish 'learning situation' (optimal conitions) from non optimal. 
  • Optimal case: presence of M in a system = perfect indicator that P is true, so the system relies on occurences of M tokens to infer that P.
  • Nonoptimal case: A system can get into a token of M with P, but the inferential mechanism is still in place.
  • Counter: what about innate representations? 
  • For example, innate perceptual processing systems.
  • Counter 2: no non circular way to distinguish 'learning situation'. Organisms can learn how to identify things without ever achieving perfection; stipulating that only situations where perfect indicator is developed is empirically risky... 
  • Cummins: It's obvious that cognitive systems can misrepresent, so cog R content cannot be a species of informational: representational content has to be defined in terms of informatonal content, without making representational content a species of informational. 
  • Representational content of M is the I content it would have under ideal conditions. This is forced on the covariantist by misrepresentation. 

Dretske's Later Theory

So, representation is not equivalent to indication. R indicates C entails if R then C, but given misrepresentaiton, R represents C does not = if R then C. Thus, Dretske suggests that representational states have the function of indicating. And, of course, functions can fail. Apparent etiological analysis of functions: states acquiring a function via selection/recrutiment.

Identifies cognitive representation with a species of functionally derived meaning.

A system S represents a property F iff S has the function of indicating (providing information about, i.e. covarying with) the F of a certain domain of objects. S performs its function by occupying different states s1, s2 ... sn, corresponding to…

Comments

No comments have yet been made