Artificial Intelligence and Human Problem Solving Processes

This paper does not draw the line between the theory of human thinking and the theory of building "thinking" machines: at the moment it makes no sense to separate them, because in this and that field of knowledge there are no concepts that are sufficiently general to explain and even more so for modeling complex intellectual activity. However, there is one difference. The fact is that psychologists working on intelligence problems have a definite tendency to reduce the number of different mechanisms involved in the models of the functioning of the human brain. This leads to attempts to achieve a greater effect with the help of fewer than the number of basic mechanisms of thinking that can be justified. Such theories do not pay enough attention to both the management of mental activity and the refinement of our knowledge about individual intellectual processes. Scientists working in the field of AI, apparently focused all their efforts on these issues, but neither of them, however, did not attach due importance to the study of the structure of knowledge, especially knowledge of the procedural type.

You can understand why psychologists do not feel very confident, using complex schemes that are not based on carefully verified mechanisms of thinking. However, the desire to limit their number still does not correspond to this stage in the development of science to the extent that this may take place in the future. Anatomy and genetics of the brain are the area of ​​knowledge in which you can assume a much larger number of different mechanisms than can be imagined today. We should focus our attention on the problems of sufficiency and efficiency, rather than on the problem of necessity.

A few years ago, the main goal of the work on image recognition was reduced to the problem of sufficiency: to find any ways leading to the development of algorithms for machine analysis of scenes. Only recently, specialists have been able to detect and realize the possibilities of properly merging individual traits and attributes into complete structures of images. I note, first of all, the works of L. Roberts (1965), A. Guzman (1968), P. Winston (1970), D. Hoffman (1971), M. Klouz (1971), J. Syray (1972), D. Waltz (1972), which characterize a series of stages in the development of questions of image-type, background-whole-part image analysis and the selection of structural groups.

Although these works are fairly simple, they can be used to give not only a superficial interpretation of the phenomenon of visual perception, but also to some extent explain the speed and smoothness of its flow. The theory of image perception faces a number of new issues in the transition from the problem of sufficiency to the problem of efficiency. How can different types of "signs" as quickly as they take place in human practice lead to the identification and description of difficult situations? What are the ways to make changes when identifying errors or finding new evidence? How are contradictions resolved? How can the information about the location of an object be changed without recomputation of the states of other related objects? What is the situation with moving objects? How do the processes of visual perception use knowledge related to common, non-visual activities? How does a person coordinate information from various sources? How can the system use expectations regarding the results of the intended actions? Can the theory explain the phenomenological results of visual perception of images, as well as the construction and manipulation of imaginary scenes controlled by the course of perception itself?

As part of the traditional approaches of behavioral and perceptual psychology, very little has been done to find answers to these questions; however, the views of some psychologists who had previously worked (see F. Bartlett, 1932), undoubtedly, were reflected in the present work. In later works on the theory of symbolic information processing, in publications similar to the articles of A. Newell (1973) and L. Pilishin (1973), there are more constructive proposals for the formulation of these controversial issues.