L#3.4.1 Reasoning Mechanisms in AI Explained | Inference Engine, Expert Systems, Rule-Based Chatbots

The fundamental tools or processes that enable an AI to “think”—that is, to make decisions, solve issues, or draw conclusions from information or rules.

1. Inference Engines

Definition:
An inference engine is a component of an artificial intelligence system that uses logical principles to infer new information from a collection of existing facts.

How it operates:

• It functions within a rule-based system similarly to a brain.
• The reasoning is as follows: B must be true if A is true AND A → B.
  B must be true if A is true AND A → B.
• It makes use of inference rules like resolution, unification, and modus ponens.

For instance,
Suppose your AI system is aware of:

• The truth is that Socrates is a man.
• All men are mortal, according to the rule.
  The Inference Engine can then infer:
• Socrates is mortal, according to this new fact.

Examples of Use:

“Use Case“ refers to real-world applications, or the situations and methods in which Inference Engines are applied in actual AI systems.

• Expert systems (such as MYCIN for diagnosing illnesses)
  • 1. Expert systems, such as MYCIN for medical diagnosis
    What is it?
    A computer program that simulates a human expert’s decision-making process is called an expert system.
    What makes it a use case?
    These systems make diagnoses or recommend treatments based on known facts (such as patient data or symptoms) and IF-THEN rules.
    For Instance:
    One of the first expert systems in medicine, MYCIN was used to identify bacterial infections and suggest antibiotics.
    How is the inference engine useful?
    It follows guidelines like this:
    If a patient has a high fever and a stiff neck, meningitis may be the cause of the infection.
• Decision-making game engines
  • 2. What are game engines for decision-making?
    AI is used in video games to give non-player characters (NPCs) intelligent behavior.
    What makes it a use case?
    The game logic uses inference engines to determine what to do next, such as whether an enemy should attack, hide, or seek cover.
    For Instance:
    An AI-controlled adversary may employ the following strategy: retreat IF the player is close AND the enemy’s health is less than 20%.
    How is the inference engine useful?
    It applies rules to act appropriately after continuously assessing game states.
• Chatbots that use logic
  • 3. Chatbots with Logic
    What is it?
    Chatbots that use logic and rules rather than just machine learning to make decisions.
    What makes it a use case?
    To determine what a user might want and how to respond rationally, they employ inference engines.
    For Instance:
    You could respond, “You can reset it at this link,” if the user says, “I forgot my password.”
    How is the inference engine useful?
    It determines the next suitable course of action by comparing user input with established facts and regulations.

Rule-based chatbots, or logic-based chatbots

• How they operate:
  They determine what response to provide using an inference engine and predefined rules (such as IF-THEN).
• Example rule:
  Respond with instructions for changing the password if the user says, "I forgot my password."
• Technology used:
  There was no machine learning utilized in the technology. entirely predicated on rules, logic, and occasionally pattern matching.

Chatbots with Machine Learning

How they function:
They use machine learning models (like neural networks, decision trees, or transformers) to predict the best response based on user input after learning from data, such as previous conversations.

Strict guidelines like “IF-THEN” are not followed by them. Rather, they pick up linguistic patterns and adjust over time.

Example:
If many users say “I can’t log in” and are responded to with “You can reset your password here,” the chatbot learns that this is the most likely correct response — even if the exact words don’t match.

Technology used:

• Natural Language Processing (NLP)
• Deep Learning
• Pre-trained models (like ChatGPT, Google BERT)

Strengths:

• Understands complex or vague language
• Learns and improves with more data
• Can personalize responses based on past interactions
  • What does “Can personalize responses based on past interactions” mean?
    It means that Machine Learning-based chatbots can adapt replies to individual users by remembering or analyzing previous conversations or behavior.
    
    🔍 Explanation:
    Instead of giving the same generic answer to everyone, ML-based chatbots can:
    Recognize repeat users
    Recall previous questions or preferences
    Adjust tone or content to suit that specific user
    
    Examples:
    E-commerce Chatbot:
    User: “Where is my order?”
    Bot: “Hi Ali, your order #12345 was shipped yesterday and will arrive tomorrow.”
    (The bot remembered Ali’s order from a past conversation or user profile.)
    Support Chatbot:
    User: “I had an issue with my last bill.”
    Bot: “Yes, I see your last issue was with overcharging. Are you facing a similar problem?”
    (It references past issues to give a smarter, relevant response.)
    
    In short:
    Personalizing responses means the chatbot remembers and adjusts replies for each user based on their previous chats, choices, or behavior.

In Summary:

• Logic-Based Chatbots are rule-driven and use inference engines (no learning involved).
• ML-Based Chatbots are data-driven, adapt with experience, and are better at handling complex, natural conversations.
• Many modern AI chatbots (like ChatGPT) combine both — logic for structure and machine learning for flexibility.