Expert Systems and AI Reasoning
In this blog post I will describe what expert systems are and discuss the different types of reasoning used by AI.
What is an expert system?
An expert system is an advanced form of Machine Learning (ML) which emulates decision making based off knowledge of human experts and advanced algorithms.
There are five components that comprise and expert system:
- Knowledge Acquisition: Knowledge is obtained from interviewing experts and then organizing the knowledge so it can be translated into programmed logic.
- Knowledge Base: This is a data repository organized into domains used by the expert system to solve problems.
- Knowledge Inference: The core of the expert system that retrieves information from the knowledge base to solve problems based off of the input.
- Explanation Module (XAI): This explains why an expert system arrived to the conclusion that it provided. This can be used for troubleshooting and fine tuning.
- User Interface: This allows for the end-user to interact with the expert system and provide input.
Types of reasoning
Complex Logic Types
A majority of AI systems use reasoning as their foundation and there are several types of reasoning used including: Inductive, deductive, Boolean, and fuzzy logic types. Inductive and deductive reasoning are described in detail in my previous post found here.
Boolean which is simply either true or false can be used in systems where the input can be something like is the car on or off. However, in some situations boolean might not represent the input well and this is where fuzzy input is used which uses input from 0-1. This is helpful to answer questions like how fast is the car moving, 0 – not at all, .5 – average driving speed, 1 – flying!
Temporal and Spatiotemporal Reasoning
Temporal reasoning deals with determining the order of events and understanding when things happen and how they can relate to each other in time. This might be used by an AI scheduling assistant to determine if you have enough time to make it from a doctors appointment to a meeting across town.
Spatiotemporal reasoning is a more complex form of reasoning that combines temporal reasoning with spatial reasoning. It deals with objects and events in both space and time. This might be used by something like an autonomous vehicle to predict the movement of other vehicles to determine if an evasive maneuver is needed.
Ontological Reasoning
Ontological Reasoning involves understanding the concepts and the relationship between them. For example, a knowledge base of animals might define concepts like Animal, Mammal, Bird, and Fish. It might then define characteristics such as the animal can fly, lives in water, or has fur. Using these properties, ontological reasoning can infer that if the animal can fly, then it has to be a bird, even if that information isn’t explicitly stated.
Taxonomy
Taxonomy is a hierarchical classification system which categorizes things based on shared characteristics. It can be visually represented by a tree diagram where each branch represents a category and then branches further into subcategories. Taxonomy is used in Biology to classify Family, Genus, and species.
Ontological Reasoning
Ontological Reasoning involves understanding the concepts and the relationship between them. For example, a knowledge base of animals might define concepts like Animal, Mammal, Bird, and Fish. It might then define characteristics such as the animal can fly, lives in water, or has fur. Using these properties, ontological reasoning can infer that if the animal can fly, then it has to be a bird, even if that information isn’t explicitly stated.
Taxonomy
Taxonomy is a hierarchical classification system which categorizes things based on shared characteristics. It can be visually represented by a tree diagram where each branch represents a category and then branches further into subcategories. Taxonomy is used in Biology to classify Family, Genus, and species.
Thanks for reading!
I hope this is helpful to somebody exploring AI. My next post will cover machine learning in more depth.
