This paper deals with an investigation of specific mechanisms of insight problem solving. We take the functioning of working memory slave system as such mechanisms. In our research we use a dual task method as a cognitive monitor (D. Kahneman) to fixing of microdynamics for study mechanisms of thinking processes. We gathered data showing that modal-specific blocks of working memory are mainly used during insight problem-solving, while solving algorithmic problems uses mainly executive control.
To date, the psychology of problem solving did not settle
the question about the legality of allocation of insight
problems in a special category. Accordingly, there are two
alternative positions. Participants of the first approach
One more method is the analysis of movements on
decision tree, used in the works of A. Newell and H. Simon
The monitoring method is based on the D. Kahneman's
uniform resource model and implies the parallel solution of
dual tasks. The monitoring method is based on the D.
Kahneman's common resource model and involves
simultaneous solution of two tasks. According to the
Kahneman's model, different structures require different
amount of attention. In addition, their resource requirements
vary in different time moments. The total amount of mental
effort (resource) that is potentially available for a system of
information processing is limited. In order to test the degree
of attention he used the method of the secondary probe task.
The main idea was that it can be possible to "catch" the
moment of the resource’s depletion when the subject deals
with the main problem (complex) and, paradoxically, cannot
cope with the additional secondary problem (simple). The
difference between total effort and an effort that is invested
in the core activity, D. Kahneman calls the spare capacity.
Spare capacity is reduced with the increasing of an effort
required to perform primary problem. Additional
(secondary) problem can only be solved using spare
capacity. If the primary problem requires more cognitive
effort, the spare capacity is reduced and the productivity of
the secondary task solution is reduced by the same amount,
and vice versa. Consequently, the change of productivity of
the secondary task reflects the change in the degree of
mental effort that is invested into primary problem.
Considering attention as a limited resource of mental effort
was further developed in theoretical and applied
investigations. Thus, D. Navon and D.Gopher
The dynamics of the resource loading on the main task is determined by the productivity of the performance of the secondary task. The primary is a thinking task (problem) and a tempo task (choice of two alternatives) performed in parallel. The reaction time and a number of errors in the implementation of the second task are fixed. The dynamics of the performance quality serves as a marker of the loading of working memory by operations which are carried out in the main problem. The secondary task must satisfy a number of requirements: a) the equal complexity, b) the same procedure of the presentation, c) the dichotomous choice, d) the equal probability of alternatives (50/50).
Decreasing productivity at a given time moment indicates that currently the resource is used by the primary task. The main problem's leading representation format (insight / algorithmic) was varied: visual or text. Similarly the type of the probe task was varied. It was required to determine the type of the angle (obtuse - acute) - visual format, or to determine the type of the syllable (open - closed) - text format. It is assumed that the coincidence of the format of the probe task and the main task creates the competition for a resource and affects the dynamics of the probe-task. To describe the dynamics and results unification the time to solve each of the problems was divided into ten equal time periods. This made it possible to unify the different time solving problems of different problems in our experimental trial.
The problem was the fact that the subject has to correct the incorrect mathematical statement by moving one match and parallel to determine the type of syllable. After the participant has identified the first syllable by pressing the corresponding key, the following syllable was presented. So until then, until the problem is resolved. In this case, we see different formats of leading representation of tasks. There are visual algorithmic problem and the textual probe task.
And at this figure, we can see the same format leading representation of tasks. There are visual algorithmic problem and the visual probe task. It’s assumed that in this case cognitive load is higher and the execution of activity is more difficult. The Stimulus material was prepared using PsychoPy - an open-source application to allow the presentation of stimuli and collection of data for a wide range of neuroscience, psychology and psychophysics experiments. It’s a free, powerful alternative to Presentation or e-Prime, written in Python (Peirce,2007)
The hypotheses are: There is a specificity of insight solution process regarding algorithmic solution; There is a domain specificity of information processing in the process of insight problem solving: the nature of the insight solution's dynamics can be reflected in the dynamics of the operation of slave systems of working memory.
1) the type of problem (insight / algorithmic) 2) the leading representation's format of the main thinking problem (visual / text) 3) the representation's format of probe task (visual / text) The dependent variable is the productivity of the probe task (reaction time)
So, the subject of this study is specific information processing in insight problem solving Participants. 58 people participated in total. (Average age is 24 years, from 18 to 56, 36 male and 22 femail. SD = 6). They were asked to solve 8 problems, but first they performed two training tasks.
1.
Investigation of the role of the central executive in solving insight and algorithmic problems was carried out by comparing the data in the performance of all tasks without focus on its representation format. An important fact is the reproducibility of the results. The graph below shows the results of two experiments performed in a uniform methodological paradigm.
There is the graph of reproducibility in two experiments. The results of the first experiment are shown in Figure 2. There is a significant dynamics in insight solving problems and its absence in algorithmic problem solving (Alg: F (9, 149)=2.3, p=.02; η =.12. Ins: F (9, 200)=1.7, p=.09; η =.07)
Similar results were obtained in our experiment (Figure 2). But in our trial all results are significant (Alg: F (9, 1479)=3.62, p<.001; η =.06. Ins: F (9, 1479)=10.12, p<.001; η =.02) (Figure 3).
We can observe that the data structure is reproduced. The greater complexity of the task in our series (more reaction time) is explained by more complexity of the probe-task (assignment to one of the two categories of 24 stimuli, whereas in the cited work the incentives were only two). We found the presence of the dynamics of working memory's loading in solving algorithmic problems. The expressed dynamics is manifested in the presence of a "hump" near the end of the solution. High productivity decreases on the steps of low-cost operations: the reading of the conditions and the voicing of the answer. Probably, the decrease in productivity near the end of the solution is connected with the implementation of combinatorial operations.
Also, one of the results of the above-cited studies (Korovkin, Vladimirov, Savinova, 2012) was the lack of a dynamics on the insight problem solving. As a possible reason was the suggestion that manipulations with a representation are important in insight problem solving and domain specific units of working memory are utilized rather than central executive. The purpose of this series was to investigate the loading of slave modal specific systems of working memory as a possible locus of information processing in the insight solutions. Separately for insight and algorithmic problems let us consider the data on the average rate performance (average reaction time) of the probe-task in conditions when its format of representation matches the format of the thinking problem and in conditions of the discrepancy between these formats.
There is a expressed cross effect (F (1, 1527) =5.96, p=.01; η=.003). When there is coincidence of the leading type of the representation of the main tasks and the probetask in the solution of insight problems then we observe a significantly lower pace of implementation of the secondary task (probe-task). Especially pronounced effect is for visual representation. A different picture is observed in algorithmic problems.
As can be seen from the graph in the algorithmic type of tasks the cross effect (F (1, 1493) =.01, p=.9; η<.001) is not expressed. Probably, these findings can be explained by the fact that in solving insight problems more important is to manipulate with the initial representation of the problem, especially for the "visual" problems where the spatial characteristics and moving into the field of the problem are important because of high uncertainty of problem space. In the type of algorithmic problems, apparently, the central executive is more important because the main resource burden falls on a phased program switch solutions that do not require taking into account the modal specifics, but it's rather more complicated switching between tasks of different modalities.
Summarizing, the data show that slave systems of working memory are less important for algorithmic problems. Central executive plays the main role and allows to keep the algorithm of solutions.
Accordingly, the model designed by H.Simon and
A.
1. A consecutive movement on the decision tree underlies the solution in the course of solving algorithmic problems. A significant role in this process is played by the central executive. For insight problems such process is not characteristic. 2. A non-directional movement in a field of the problem (obviously, it's a spatial or quasi-spatial structure) underlies the solution in the process of insight solving problems.
This work is supported by Russian Foundation for Basic Research (grant 15-06-07899a) and The Mikhai Prokhorov Foundation (Karamzin grant program 2015)