Reasoning: The Pillar of Artificial Intelligence

Reasoning: The Pillar of Artificial Intelligence

Reasoning refers to the logics or techniques being used by Intelligent agents or Expert systems to represent the knowledge what they got by different types of learning like supervised, Non- supervised, Reinforcement or Semi- supervised learning.

Before moving in depth of the reasoning part, let us know about the different ways by which one can know something or collect information for further providing those details for training purpose of intelligent agents. As we know we or humans make intelligent agents by providing training data to it. For that purpose we need data.

So, following are various ways of knowing something or collection of information:

• Empirical method
• Scientific Method
• And many more

Now, we should know that what is reasoning? Reasoning can be understand as inferences. If an agent knows something, what else can that agent infer from the given facts.

There are 3 types of inferences:

1. Deductive inference
2. Abductive inference
3. Inductive inference

• Deductive inference means inference drawn would be necessarily true. For example, if a patient have Typhoid, then the inference arouses from this fact is that the patient must have fever. This is a necessarily true fact inferred from the given fact.

This type of inference is also called as cause to effect .

• Abductive inference means inference drawn will be possibly true. For example, if a patient is having fever  and went to doctor for further diagnosis, then possible inferences might be Typhoid, Dengue etc.

This type of inference is also called as effect to cause.

• Last type of inference is known as Inductive inference. When we generalizes the fact as an inference from the given facts, then inductive inference takes place.

For example-

1. Peepul leaf is green.
2. Tamarind leaf is green.
3. Neem leaf is green

Inductive inference will be “All leaves are green” 

When the facts inferred, then the role of logic came for Reasoning in artificial intelligence. Logic is a branch of mathematics used to validate any predicate or relationship between individual elements by using  truth tables. It plays a prominent role in reasoning  by proving any sentence to be valid or not. For example, if any sentence is proven as tautology then that sentence is valid otherwise not.

There are 2 types of sentences:

1. Atomic or primitive sentences
2. Compound sentences or we can say that complex sentences which are formed by the use of logical connectives.

Logical Connectives are used to describe new knowledge by taking one or more than one sentence as input.

Mostly used Logical connectives are:-

1. Unary – It takes one Sentence as input. Example- logical NOT
2. Binary connectives- It takes two Sentence as input. Example-

∧ usually read as AND. If A and B are sentences then so is (A ∧ B)

∨ usually read as OR. If A OR B are sentences then so is (A ∨ B)

⊃ or → or ⇒ usually read as IMPLIES. We will use the first one.

If A then B  or All A’s are B is the sentence then it will be represented as (A ⊃ B)

… there are 16 binary connectives

1. Ternary connectives – It takes 3 sentences as input. But generally ternary or Higher Order Connectives are not required. It is efficiently expressed as a combination of Unary or Binary Connectives.

Author Name: Ms. PRATIMA KUMARI

Department: MCA

Email Id: pratimak.phd22.cs@nitp.ac.in