Runeson argues that perception consists of 'Smart' Perceptual Mechanisms. He compares them to the polar planimeter, an instrument which allows for the measurement of areas on a 2D surface such as a map. These smart mechanisms are similar to the affordances that Gibson sees as the basis for perception. Runeson does not see the need for cognitive processes in perception. As in the planimeter, the relationship between the stimulus and the smart mechanism is automatic, emanating from the ‘physical realisation’ of the mechanism.
This is fine if the phenomenon under study is perception and the unit of analysis is the smart mechanism. But if we want to dig lower, how does it work? The distance the planimeter travels in any direction is directly related to the area covered by the arm. A mathematical proof is available. Can smart mechanisms be explained at a lower level?
The smart mechanisms emerge from lower level processes that we can study. We know that light hits the retina and is imprinted on to the visual cortex. What happens then? How do the smart mechanisms emerge? It would seem that some sort of processing must take place. There is some algorithm. This is not saying it is digital, but it sure is not the same as in the planimeter, as there are not any arms or wheels involved. We have neurons that are activated in direct relation to the object being perceived. The process allows us to categorise and learn. What happens to give us smart perception? It’s OK to skip the activation of the neurons and consider smart mechanisms and affordances, as long as it is acknowledged that this level is being skipped.
If the logical process is not digital or analogue, maybe it’s bio-logic. It is driven by data from our senses. It supports all the processes from the lowest level of imprinting light on the neurons, to smart mechanisms and the higher level conscious processes such as attention and emotion. Higher level processes emerge from lower level processes in some extremely complex manner.
During development, data is experienced and allows discovery of invariants such as Runeson describes: The time to collision is related to the rate of expansion of the image on the eye. The time to jump a fence depends only on the height of the fence. Complex dynamic systems, neural networks and machine learning are data based systems as opposed to rule based. They show the ability to find hidden and latent relationships in data. These appear similar to invariants in that they are not based on rules or theories, but just on the nature of the data.
The planimeter differs from perception in that the intelligence it displays did not emerge from it, but was designed into it, whereas smart perceptual mechanisms emerge. Runeson’s analysis of the planimeter allows us to understand that laws can be implemented in other ways than in formal disciplines such as mathematics and physics. It allows us to appreciate that there is a logic in the biological that does not have to be the same as the logics with which we are more familiar.