Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:

---

Introduction
Today, various cognitive models that represent the brain’s cognitive processes in areas such as psychology and artificial intelligence have been proposed that have found many applications. According to the objectives of these models, the most important of which is to study the characteristics of the brain in the process of performing excellent cognitive functions, rehabilitation of patients with cognitive impairments and intelligent machines, it is necessary to review and compare the proposed models carefully. Given that there are different cognitive functions, this research have considered these three models more based on the capabilities and models provided for memory, learning, and perception that are the core of cognitive functions. These models have been examined independently in research often before the capabilities of the models are studied in order to compare them. The present study will also point out the new and unique capabilities that the free energy model of the brain provides compared to other models. It is essential to note that, more or less, most models can be used in both the cognitive and artificial intelligence domains. However, the capabilities of these models differ in the applications they have for each domain. So that, for example, the model of semantic networks in terms of application in the field of artificial intelligence remains on the same premise In contrast, the free energy model of the brain has the ability to use various applications in artificial intelligence and cognitive psychology.
Methods
In this research, information related to three types of cognitive models was examined and compared by collecting library data. Among the models proposed by cognitive psychologists are the semantic network model, the ACT-R model, and the free energy model of the brain, which have been studied and compared here. In the process of producing news knowledge, information that is arranged in a hierarchy or relationship based on meaning or common features in semantic memory and  the form of a network consisting of concepts and communication links between them (semantic network) will produce declarative knowledge. Thus, semantic knowledge results from input perception and agent’s inference from these perceptions, which is stored in semantic memory as a communication network. This network causes semantic organization, which allows the agent to retrieve information. In the process of producing procedural knowledge, a set of possible actions is required to achieve a specific goal. This method is known for understanding the process of procedural learning as a pattern of sequential information processing. The ACT-R model combines some features of semantic networks and some features of sequential processing patterns and is presented in the form of a model based on the relationship between memory types. In this model, the relationship between declarative, production (procedural), and working memories forms the basis of cognitive processes. Environmental states cause different stimuli, and the presence of stimuli in the environment indicate hidden or revealed states. On the agent side, after receiving environmental stimuli that may have been subject to uncertainty or various changes until the agent perceived, there are a set of uncertain or probable perceived stimuli indicates the existence of states or concepts. Receiving sensory data and increasing entropy is a surprise. To avoid brain disorder, the surprise that results from inconsistent data with hypotheses in the brain must be dispelled immediately through free energy consumption. Accordingly, the free energy principle (FEP) shows that any adaptive change in the brain will minimize free energy. Two processes of perception and action do this minimization. Perception means changing expectations to reduce entropy and prediction error, and action means changing the agent’s configuration by affecting the biological agent in the environment in order to change the sensory stimuli and to avoid surprise or surprise. Combining perception and action makes it possible to adapt to new sensory stimuli. This process is called active inference.
Results
The model of semantic networks is based on declarative memory, which allows the generation of semantic knowledge. The ACT-R model, one of the most practical and up-to-date models for examining human cognitive characteristics, has been proposed based on declarative and procedural memories, enabling the generation of declarative and procedural (skill) knowledge. Models of semantic networks and free energy require programming, while the ACT-R model is presented in the form of flexible and user-friendly software. The free energy model of the brain works very similarly to human inferences by generating a variety of knowledge that updates concepts and perceptions based on probabilistic inferences by Bayesian updates of previous brain hypotheses and by minimizing brain free energy. In this study, a proposal is made to use the free energy model of the brain and turn it into a model for the generation of declarative, procedural, and conditional (a combination of both declarative and procedural knowledge). This model has the ability to generate concepts and construct propositions based on the concepts produced and the relationship between them as a human agent. It can also teach procedural knowledge such as how to drive. Also, it can combine declarative and procedural knowledge to solve conditional problems such as solving mathematical problems. Although the ACT-R model also has the ability to generate procedural and declarative knowledge, due to insufficient ability to model perceptual similar to a human, it has limitations in the production of these two types of knowledge in the production of conditional knowledge. The two models of free energy and ACT-R offer the possibility of better explaining the different processes that lead to cognitive diseases. Also, by modeling diseases by these models, predictions can be made about cognitive rehabilitation. The free energy model shows some cognitive diseases more accurately so that in diseases such as schizophrenia, attention disorders, Alzheimer's, and memory impairments, it can suggest practical solutions to control or reduce each.
Conclusion
The present study examined three cognitive models of semantic networks, ACT-R, and the free energy of the brain. The capabilities of these models in various fields of cognition, artificial intelligence, cognitive rehabilitation, and modeling of cognitive functions were considered. Finally, it can be said that the free energy model has many advantages in modeling the human brain compared to other models. In future work, while using two models of free energy and ACT-R in areas such as diseases, significant progress can be seen. In diseases such as schizophrenia and attention deficit disorder, according to the inferential model presented by the free energy model of the brain, good success has already been achieved. Despite the complexities of the free energy model,  due to its more significant comprehensiveness, in providing a more accurate and calculated explanation of various cognitive functions in the brain such as perception, learning, attention, decision making, and more appropriate analysis of human cognitive diseases, this model can give much broader and more accurate results.
Ethical considerations
Compliance with ethical guidelines
This study was performed in a library method without participants or other people. There was no characteristic of not observing moral principles in interaction with people.
Funding
No financial support was received from any organization in conducting this study.
Author’s contributions
The corresponding author prepared this article, and other authors have contributed as a supervisors or advisors.
Acknowledgments
The authors thank Professor Karl Friston, professor at the University College of London, for his guidance, submission of articles, and the results of his research, which have been very influential in this study.
Conflict of interests
There was no conflict of interest with any natural or legal person in this study.

Keywords: Semantic networks, ACT-R, Free energy model, Active inference

Full-Text [PDF 1204 kb]   (449 Downloads)    

Add your comments about this article : Your username or Email:

---

---