Elements of a cybernetic epistemology: elementary anticipatory systems

Abstract

Purpose -The purpose of this paper is to analyze how elementary anticipation, understood as anticipation of the repetition of one known pattern, can emerge out of sequence learning and how it can contribute to the behavioral options of goal-oriented systems. Design/methodology/approach -A functional approach is used to develop the necessary cybernetic structures of a subsystem for sequence learning that can additionally provide standards of anticipated patterns for future pattern matching. Based on that it is analyzed, how a goal-oriented system can use the information about the actual occurrence of an anticipated pattern. Findings -A subsystem for elementary anticipation of single patterns builds on sequence learning and requires additionally a structure: first, to unequivocally identify the beginning of known sequences just from their first patterns; and second, to decide to use a latter pattern of such a sequence as standard for an anticipated pattern. Deciding to actually use such a pattern for anticipation requires an additional subsystem to switch between the feedback pattern recognition mode and feedforward. Then the occurrence of such an anticipated pattern allows immediate recognition and action.Practical implications -The paper shows a necessary evolution of cybernetic structures from pattern recognition via sequence learning to anticipation; and it shows, too, a necessary evolution in the cognitive development of individual systems. In the simple anticipatory structures analyzed here, only known patterns, that are part of a known sequence, can become anticipated patterns. Originality/value -The paper places elementary anticipation of single patterns in an evolutionary development based on pattern recognition and sequence learning. It provides the base to analyze more complex forms of anticipation.

Elements of a cybernetic epistemology: complex anticipatory systems

“…We can show again that the content of short-term memory determines the behavioral options of a goal-orientated system at a point in time. We showed that already for pattern recognition and sequence learning systems (Nechansky, 2012a, b) as well as for elementary anticipatory systems (Nechansky, 2013). All these system can momentarily only act towards patterns temporarily held in short-term memory, while any other observed patterns cannot be recognized and therefore cannot serve as inputs in decision-rules to trigger actions.…”

“…We will argue below that this decision process can use the same bottom-up and top-down search to find candidates for anticipation and the same checking for evaluations, as we found for elementary anticipation (Nechansky, 2013). But here it requires repeated cycles of this searching and checking, till finally the most important (dangerous or useful) pattern can be selected for anticipation.…”

“…Here the same criteria apply as for elementary anticipation (Nechansky, 2013): anticipation has to end:…”

“…In principle this decision process is equal with the initiation of elementary anticipation (Nechansky, 2013). But it requires either a very fast serial or a highly parallel structure to check at one point in time for the evaluation of more patterns, movements, etc.…”

“…As for elementary anticipation we suppose here a structure that can keep anticipated patterns in short-term memory 2, while using short-term memory 1 for feedback pattern recognition in case the anticipated pattern does not show (for the details see Nechansky, 2013).…”