Chapter

Unconscious Inference Theories of Cognitive Achievement

Authors:
To read the full-text of this research, you can request a copy directly from the authors.

Abstract

This chapter argues that the only tenable unconscious inferences theories of cognitive achievement are ones that employ a theory internal technical notion of representation, but that once we give cash-value definitions of the relevant notions of representation and inference, there is little left of the ordinary notion of representation. We suggest that the real value of talk of unconscious inferences lies in (a) their heuristic utility in helping us to make fruitful predictions, e.g., about illusions, and (b) their providing a high-level description of the functional organization of subpersonal faculties that makes clear how they equip an agent to navigate its environment and pursue its goals.

No full-text available

Request Full-text Paper PDF

To read the full-text of this research,
you can request a copy directly from the authors.

ResearchGate has not been able to resolve any citations for this publication.
Article
It is a mistake to consider perception and learning separately because what one learns is strongly constrained by what one perceives, and what one perceives depends on what one has experienced. I shall propose the hypothesis that perception is the computation of a representation that enables us to make reliable and versatile inferences about associations occurring in the world around us--that is, perception prepares the ground for learning. The statistical problem in learning is to determine whether a compound event such as "C followed by U" is a random co-occurrence or a significant association, for if it is the former it would be a mistake to pay any particular attention to C, whereas if it is the latter C is a conditional stimulus for U and a useful predictor for it. Now you cannot decide whether the association is random or not without knowledge of the prior probabilities of C and U: hence on my hypothesis when you perceive an object or event the representation must not only signal "it's there" or "it's happened", but must also make evident (or rapidly accessible) the prior probability of what has been signalled. Furthermore it must do this for all the objects or events that can act as conditional stimuli, and this implies that the representative elements should be statistically independent (or approximately so) in the normal environment. Forms of coding that would do this, and the relationship with Helmholtz's unconscious inference, will be discussed. These considerations imply that the task performed in perception has been overlooked both by learning theorists and by connectionists working on associative and adaptive networks. Coding for independence may be particularly important in understanding the developmental processes during the sensitive period: it may be the operation that leads ontogenetically-timed, activity-dependent, connections to imprint appropriate codes if the animal has experience, but inappropriate codes without experience.