31.5 Cognitive and connectionist nets

For some decades now, cognitive psychologists have modeled mental processes in terms of cognitive or connectionist networks. These have often been formally defined in terms of semantic associations. My question now is whether these semantic networks can be re-worked to explicitly derive the manner of existence of ideas or thoughts in some kind of associative space. What kind of metric would be implied for such an associative space?

Part of the answer here depends on which kinds of operations are needed describe cognitive processing. Operations are the modification of those ideas and those semantic connections. Very often these are defined in terms of algorithmic procedures attached within the semantic net itself. From the viewpoint of theistic science, however, these procedures should be taken as approximate treatments of the dispositional character of the several dispositions then present and active. Section 22.2 has discussed how the theistic science ontology may give rise to functions and procedures. Now we are asking whether existing procedural modeling in cognitive psychology could be mapped back onto ontologies of dispositions.

Most existing work on cognitive nets involves what theistic science would regard as a single level and perhaps also the neighboring levels in order to implement procedures and memories. To understand the general operation of multilevel structures, we could look for guidance at the work of Pacherie (2008), who draws up the flow diagrams of intention, action control and feedback involving at least three levels of psychological processing. Suggestions have been made to generalize the structure of cognitive or connectionist networks. Thompson (1990) has talked about ‘layered networks’, and Commons (2008b) has talked about ‘stacked neural networks’, but neither has yet been implemented or tested.