Sabrina presented an invited talk this week at the Computation & Representation in Cognitive Science conference at the University of Sussex
Ecological representations: Can the ‘R’ word fit in a Gibsonian framework?
There is widespread agreement that an ecological approach to explaining behaviour is at odds with a computational approach. Often, this opposition is also framed in terms of representations such that ecological approaches are meant to be necessarily non-representational. The perceived lack of fit between ecological explanations and computational / representational explanations may be accounted for by two facts. One, cognitive scientists tend to adopt an unnecessarily narrow view of computation (compared to the wider scientific community). Two, the motivation for invoking representations in an explanation is unfairly reduced to solving a problem of poverty of stimulus, which is anathema to ecological explanations. We argue that the concept of representation is amenable to an ecological approach, as long as it is built upon a foundation of ecological information. Ecological representations do not fill out impoverished sensory experiences. Their usefulness comes entirely from the extent to which they preserve spatiotemporal structure in information variables specifying biologically and psychologically relevant properties of the environment. Their job is to provide a mechanism by which behaviour can complement relevant properties of the environment in the absence of immediate perceptual access to those properties. Because the structure of ecological representations is determined by the structure of ecological information (which is formally definable), ecological representations are amenable to empirical investigation, making their existence an empirical, rather than theoretical matter.