top of page

Representing human knowledge in AI systems

1 March 2023
Zvia Hen
Girl speaking with robot

The essential difference between humans and computers is human wisdom, which is based on advanced knowledge and experience we have gained in our lives that allow us to think, make decisions and act in a certain way.

Artificial Intelligence systems are based on innovative technology, which reflects the power of human reason combined with machine computers.

Types of human knowledge
  • Declarative Knowledge - describing the world: ideas, facts, concepts.

  • Structural Knowledge - basic knowledge, which describes the system of connections between concepts, ideas, and their description. For example, part of..., belongs to a group, is similar to...

  • Procedural Knowledge - Laws, strategies, processes, agendas.

    Apt conduct or operation in a particular situation. The practical knowledge needed to perform a task.

  • Meta Knowledge- knowledge about the other types of knowledge.

  • Heuristic Knowledge- experience-based knowledge held by experts in specific fields and subjects. Rules of thumb, practices, and recommendations. This knowledge can be considered "tacit knowledge" that the learner will use when solving problems and making decisions.

Correspondingly, types of knowledge in artificial intelligence systems:
  • Object - all the information, characteristics, and facts about an object. For example, a bus has wheels; a guitar has strings.

  • Event - events happening in our world.

  • Action - an explanation of human behavior action in a particular situation.

  • Meta knowledge - knowledge about the other types of knowledge.

  • Facts - truths, proven knowledge about the reality we live in and our beliefs.

  • Knowledge base - all the information on a field or discipline. For example: road construction.

What is knowledge representation in artificial intelligence systems?

The concept of knowledge representation in artificial intelligence systems is based on the transfer of advanced human knowledge to these systems.

Knowledge representation in artificial intelligence systems involves the following aspects:

  1. Characteristics of human thinking and a thinking model that contributes to intelligent behavior of the system.

  2. Representing information about the world in a way that a computer can understand and apply to solve complex problems. For example, diagnoses of medical conditions or the ability to communicate with humans in human language.

  3. Learning from data to develop human intelligence.

There are four main approaches to representing knowledge in artificial intelligence systems:

  1. Relational Knowledge- sets of data on the object, represented in columns. This approach is common in databases that represent relationships between entities. This approach leaves almost no room for drawing conclusions.

  2. Inheritable knowledge - data stored according to hierarchies of categories. The knowledge represents the connections between an "event" and a category. The elements draw values according to the hierarchy.

  3. Inferential Knowledge - this approach represents logic-based knowledge and is used to create additional facts, and is considered accurate.

  4. Procedural knowledge - using programs and codes that describe how to perform specific actions. One of its important rules is: if...then. Implementing this approach, you can use a variety of programming languages, such as LISP and Prolog.

To reflect intelligent human behavior, an Artificial Intelligence system should include the following components:

  • Perception – the data can be retrieved via a visual, audio or other sensory input.

  • Learning - learning from the data inputted to the system through the retrieval component.

  • Knowledge Representation and Reasoning - represent a quasi-human intellectual activity.

  • Planning - depends on analyzing the representation of knowledge and conclusions.

  • Education- depends on the analysis of the knowledge representation and conclusions.

What are the characteristics and requirements for a good artificial intelligence system?
  1. Accurately representing all types of knowledge required in a particular field.

  2. Conclusion accuracy - systemic flexibility, the ability to deal with all types of existing knowledge, manipulate it, and create new knowledge based on existing knowledge.

  3. Inferential efficiency - an effective mechanism for assimilating new knowledge.

  4. Effective knowledge acquisition - acquiring new knowledge using automatic methods.

The optimal development of artificial intelligence systems depends on the ability to endow them with human wisdom. The key lies in sound knowledge representation systems, which will enable knowledge to be imparted to AI systems the same way knowledge is imparted to humans.


What is knowledge representation?

Knowledge representation in Artificial Intelligence

bottom of page