M. Ya. Rozenberg

Nowadays, consideration of human factor impact on software development leaves much to be desired, i.e. human factor in modern estimation models is represented as some coefficients of abstract nature (e.g. personnel skills, personnel motivation, etc.). In practice such human factor representation in estimation models results in negative consequences like missed software development deadline, its cost overrun, quality decreasing, and even project development abortion. In formulation procedure it is reasonable to rely on verified data in neurophysiology and artificial neural networks because many human factor aspects are closely connected with the mode of human thinking. In this work formal description of various human factor quality properties is suggested, qualification, teaching, forgetting, information perception limits etc. being among them. Formal descriptions given are based on extended neuron model. It is proposed that, firstly, there exist output signal transmission mechanisms acting not only in the direction from axon to dendrite but also backwards; secondly, output signal can be divided among separate dendrite branches and an axon in arbitrary proportion. The extensions mentioned can represent functional roles of an axon and dendrites as follows: dendrites bind the entity presented by a neuron with its parts or particular cases (entities of lower abstraction), and that of axon - with entities of higher abstraction. As qualification in its essence is a sustainable set of knowledge on splitting entities into abstraction tiers, the model used allows representing qualification as neuron interactions. Characteristic feature concerning the mode of output signal distribution between an axon and dendrite branches in the model considered allows information perception limits formalizing in task decomposition procedure, this giving the possibility to introduce mathematically sound criteria for task complexity. Besides, formal description of a number of auxiliary notions (teaching, forgetting, attention concentration on sub-tasks etc.) is given in the article under consideration; these notions can be represented through other characteristics of the model applied. Human factor aspects models submitted are considered to be essential ones in relation to software development procedure, and are in good agreement with empiric data known. Although the above mentioned models are designed mostly for qualitative, but not quantitative results, nevertheless, they can serve as a basis for calculation models creation, those being widely applied in software development practice.