Engineering Verification of Cognitive Models: A Software Simulation Approach

The construction of intelligent agents in accordance with the philosophy, which is expressed by the metaphor “a system that thinks like a person,” involves the use of models of various systems and phenomena developed by cognitive psychologists. However, the use of cognitive models in the synthesis of intelligent agents is difficult for several reasons. Firstly, descriptions of cognitive models contain a significant amount of uncertainty, since they are based mainly on the use of natural language. Secondly, the cognitive models described in the literature are usually not verified from an engineering point of view. By engineering verification of a cognitive model, we mean the development of a software simulator of the model with the help of which it is possible to conduct those experiments that were previously carried out in psychological experiments with subjects. If the results of experiments conducted using the simulator match the results of the original experiments with subjects, then the model can be said to have passed engineering verification. The article describes the process of engineering verification of the well-known cognitive model of focused auditory attention, which was proposed by Broadbent. At the first stage of the study, a formal version of Broadbent's model of attention was developed. At subsequent stages, the formal version of the model was used to develop a software simulator. At the final stage of the study, a software simulator was used to conduct experiments on dichotic listening to binaural lists. The results of the experiments showed almost complete agreement between the correct answers obtained in psychological experiments and using the simulator.