Reasoning
Reasoning in AI refers to the process by which a system derives new conclusions from existing knowledge. There are three classical forms. Deductive reasoni...
Forms of Logical Reasoning
Reasoning in AI refers to the process by which a system derives new conclusions from existing knowledge. There are three classical forms. Deductive reasoning starts with general premises and derives specific, guaranteed conclusions. If all mammals are warm-blooded and a dolphin is a mammal, then a dolphin must be warm-blooded. Deduction is truth-preserving: if the premises are true and the reasoning is valid, the conclusion cannot be false.
Inductive reasoning moves in the opposite direction, drawing general conclusions from specific observations. Observing that the sun has risen every day for thousands of years leads to the general claim that the sun rises every day. Induction is the basis of machine learning: a model observes many training examples and generalises to unseen cases. However, inductive conclusions are not guaranteed to be true; they are probabilistic.
Abductive reasoning selects the simplest or most likely explanation for an observation. A doctor seeing a patient with fever, cough, and fatigue abduces that the patient probably has a respiratory infection rather than a coincidental combination of unrelated conditions. Abduction drives diagnostic AI systems. In addition to these classical forms, non-monotonic reasoning handles situations where new information can overturn previous conclusions, which is essential for real-world AI because the world changes and knowledge is rarely complete.
Forward and Backward Chaining
In rule-based systems, forward chaining starts from known facts and applies rules to derive new facts until the goal is reached, working data-driven. Backward chaining starts from the goal and works backward to find supporting facts, working goal-driven. PROLOG uses backward chaining as its default inference mechanism.
Exam Tip
A classic exam question gives a knowledge base of facts and rules and asks you to trace forward or backward chaining to prove or disprove a goal. Practise writing out each inference step explicitly. Also know that modus ponens, the rule that if P implies Q and P is true then Q is true, is the most fundamental deductive inference rule you will encounter.