Artificial Intelligence - What Are Cognitive Architectures?


A cognitive architecture is a customized computer model of the human mind that aims to imitate all elements of human cognition completely.

Cognitive architectures are coherent theories that explain how a set of fixed mental structures and mechanisms may do intelligent work in a range of diverse surroundings.

A cognitive architecture has two main components: a theory of how the human mind works and a computing representation of the theory.

The cognitive theory that underpins a cognitive architecture will attempt to bring together the findings of a wide range of experiments and hypotheses into a single, comprehensive framework capable of describing a wide range of human behavior utilizing a set of evidence-based processes.

The framework established in the theory of cognition is then used to build the computational representation.

Cognitive architectures like ACT-R (Adaptive Control of Thought Rational), Soar, and CLARION (Connectionist Learning with Adaptive Rule Induction On-line) can predict, explain, and model complex human behavior like driving a car, solving a math problem, or recalling when you last saw the hippie in the park by combining modeling behavior and modeling the structure of a cognitive system.

There are four techniques to reaching human-level intelligence inside a cognitive architecture, according to computer scientists Stuart Russell and Peter Norvig: 

(1) constructing systems that think like people, 

(2) constructing systems that think logically, 

(3) constructing systems that behave rationally, and 

(4) constructing systems that act like humans.

The behavior of a system that thinks like a person is produced using recognized human processes.

This is the most common technique in cognitive modeling, as shown by structures such as John Anderson's ACT-R, Allen Newell and Herb Simon's General Problem Solver, and the first applications of the general cognitive architecture known as Soar.

For example, the ACT-R model combines theories of physical movement, visual attention, and cognition.

The model differentiates between declarative and procedural knowledge.

Production rules, which are statements expressed as condition action pairs, are used to express procedural knowledge.

A statement written in the form of IF THEN is an example.

Declarative knowledge is grounded on reality.

It refers to data that is considered static, such as characteristics, events, or objects.

This sort of architecture will provide behavior that contains both right and incorrect behavior.

Instead, a system that thinks rationally would employ logic, computational reasoning, and rules of mind to create consistent and correct behaviors and outputs.

A rational system will employ intrinsic ideas and knowledge to attain objectives via a more broad logic and movement from premises to consequences, which is more flexible to circumstances when complete information is not available.

The rational agent method is another name for acting rationally.

Finally, the Turing Test technique may be viewed of as creating a system that behaves like a person.

To attain humanlike behavior, this technique requires the development of a system capable of natural language processing, knowledge representation, automated reasoning, and machine learning in its most stringent form.

This method does not require every system to achieve all of these requirements; instead, it focuses on the standards that are most important to the job at hand.

Apart from those four methods, cognitive architectures are divided into three categories based on how they process information: symbolic (or cognitivist), emergent (or connectionist), and hybrid.

Symbolic systems are controlled at a high level from the top down and analyze data using a series of IF-THEN statements known as production rules.

EPIC (Executive-Process/Interactive Control) and Soar are two examples of symbolic information processing cognitive systems.

Emergent systems, like neural networks, are constructed by a bottom-up flow of information propagating from input nodes into the remainder of the system.

Emergent systems like Leabra and BECCA (Brain-Emulating Cognition and Control Architecture) employ a self-organizing, distributed network of nodes that may function in parallel.

ACT-R and CAPS (Collaborative, Activation-based, Production System) are examples of hybrid architectures that integrate elements from both forms of information processing.

A hybrid cognitive architecture geared at visual perception and understanding, for example, may utilize symbolic processing for labels and text but an emergent method for visual feature and object recognition.

As particular subtasks become more known, this kind of mixed-methods approach to developing cognitive architectures is becoming increasingly widespread.

This may lead to some confusion when categorizing designs, but it also leads to architectural improvement since the best solutions for each subtask can be included.

In both academic and industry settings, a number of cognitive architectures have been created.

Because of its sturdy and speedy software, Soar is one of the most well-known cognitive architectures that has gone out into industrial applications.

Digital Equipment Corporation (DEC) utilized a proprietary Soar program named R1-Soar to help with the complicated ordering process for the VAX computer system in 1985.

Previously, each component of the system (from software to cable connections) would have to be planned out according to a complicated set of rules and eventualities.

The R1-Soar system used the Soar cognitive architecture to automate this operation, saving an estimated $25 million each year.

Soar Technology, Inc. maintains Soar's industrial implementation, and the business continues to engage on military and government projects.

John Anderson's ACT-R is one of the most well-known and researched cognitive architectures in academia.

ACT-R builds on and expands on previous architectures such as HAM (Human Associative Memory).

Much of the research on ACT-R has focused on expanding memory modeling to include a wider range of memory types and cognitive processes.

ACT-R, on the other hand, has been used in natural language processing, brain area activation prediction, and the construction of smart tutors that can model student behavior and tailor the learning curriculum to their unique requirements.

~ Jai Krishna Ponnappan

You may also want to read more about Artificial Intelligence here.

See also: 

Intelligent Tutoring Systems; Interaction for Cognitive Agents; Newell, Allen.

Further Reading

Anderson, John R. 2007. How Can the Human Mind Occur in the Physical Universe? Oxford, UK: Oxford University Press.

Kotseruba, Iuliia, and John K. Tsotsos. 2020. “40 Years of Cognitive Architectures: Core Cognitive Abilities and Practical Applications.” Artificial Intelligence Review 53, no. 1 (January): 17–94.

Ritter, Frank E., Farnaz Tehranchi, and Jacob D. Oury. 2018. “ACT-R: A Cognitive Architecture for Modeling Cognition.” Wiley Interdisciplinary Reviews: Cognitive Science 10, no. 4: 1–19.

What Is Artificial General Intelligence?

Artificial General Intelligence (AGI) is defined as the software representation of generalized human cognitive capacities that enables the ...