Premium
Estimating Human Cognitive Capacities: A Response to Landauer *
Author(s) -
Hunter Lawrence
Publication year - 1988
Publication title -
cognitive science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.498
H-Index - 114
eISSN - 1551-6709
pISSN - 0364-0213
DOI - 10.1207/s15516709cog1202_5
Subject(s) - reading (process) , citation , cognitive science , cognition , psychology , computer science , library science , philosophy , linguistics , neuroscience
Thomas K. Landauer’s (1986) estimate of the capacity of normal human memory is deeply flawed. His estimate neither reflects the quantity of computational resources necessary for supporting a human-like memory, nor provides any constraint on a theory of such a memory. The roots of the problem lie in a m isunderstanding of the relationship of previous experience to intelligent behavior and therefore a failure to provide the kind of estimate that could be useful for constructing an AI or psychological theory of remembering. The word memory has two quite distinct meanings. One refers to the accumulated experience of a human being; it is the difference between a naive youth and a wise elder. Human memory is not confined to the ability to recognize previously seen objects or accurately reproduce previously seen texts. It is our memories that allow us to recognize and refine categories of objects or events, to-generate reasonable expectations and plans, or to focus attention on salient aspects of the environment, to name but a few of the tasks memory subserves. In short, human memory is the ability to bring previous experiences to bear on new situations. The other meaning of memory is the specifically defined term of information theorists, most commonly encountered in discussions of computers; that is, a device that can maintain one of several possible states as a result of past action, thereby preserving information over time. The smallest unit of memory is the smallest unit of information, or a bit. Any system whose state depends on some past event can be characterized in information theoretic terms, and the amount of information it can transfer from past events to current state can be measured in bits. Landauer’s goal seems to be to estimate the number of bits it would take to duplicate the store of experiences of a human being in a way that subserves the same functions as human memory. As Landauer acknowledges, the relationship between human-like memory and the underlying computer-memory requirements necessary to support it