Topic 1: Knowledge-Based Agents

What is a Knowledge-Based Agent?

A Knowledge-Based Agent (KBA) is an intelligent agent that uses knowledge about its environment and logical reasoning to make decisions, rather than relying solely on reactive responses to stimuli.

Key Characteristics:

Knowledge Storage: Maintains a knowledge base (KB) containing facts and rules
Reasoning Capability: Can derive new information from existing knowledge
Goal-Oriented: Uses reasoning to achieve specific objectives
Adaptive: Can learn and update its knowledge base

KBA vs Reactive Agents

Aspect	Reactive Agents	Knowledge-Based Agents
Decision Making	Direct stimulus-response	Reasoning-based decisions
Memory	No persistent memory	Maintains knowledge base
Learning	Cannot learn from experience	Can learn and update knowledge
Complexity	Simple, fast responses	Complex reasoning processes
Flexibility	Limited to programmed responses	Highly flexible and adaptable
Example	Thermostat, simple robot	Expert system, chess program

The Role of Knowledge Base (KB)

The Knowledge Base is the heart of a knowledge-based agent. It stores facts and rules about the world that the agent can use for reasoning.

What's Stored in KB:

Facts: Basic statements about the world
Rules: If-then relationships
Axioms: Fundamental truths
Constraints: Limitations and boundaries

KB Operations:

Tell: Add new knowledge
Ask: Query existing knowledge
Retract: Remove outdated knowledge
Update: Modify existing knowledge

Example Knowledge Base:

// Facts
is_bird(tweety)
has_wings(tweety)
can_fly(X) :- is_bird(X), has_wings(X)

// Rules
if can_fly(X) then is_animal(X)
if is_animal(X) then is_living(X)

The Sense-Reason-Act Cycle

Knowledge-based agents follow a continuous cycle of perceiving, reasoning, and acting:

Perceive

Tell

Ask

Act

1. Perceive

Agent observes the environment through sensors

2. Tell

Add new percepts to the knowledge base

3. Ask

Query KB to determine what action to take

4. Act

Execute the chosen action in the environment

The Declarative Approach

The declarative approach focuses on what is true rather than how to act. This separation of knowledge and control provides several advantages:

Advantages:

Modularity: Knowledge and control are separate
Reusability: Same knowledge can be used for different tasks
Maintainability: Easy to update knowledge without changing code
Transparency: Knowledge is explicit and understandable

Challenges:

Representation: Need formal languages for knowledge
Inference: Requires efficient reasoning algorithms
Completeness: Must represent all relevant knowledge
Consistency: Knowledge must not contradict itself

PEAS Examples: Wumpus World

The PEAS framework helps us analyze agent design by considering Performance, Environment, Actuators, and Sensors.

Wumpus World Agent:

Performance:

Find gold
Avoid wumpus
Avoid pits
Minimize steps

Environment:

4×4 grid world
Partially observable
Stochastic (wumpus location)
Sequential

Actuators:

Move forward
Turn left/right
Grab gold
Shoot arrow

Sensors:

Stench (near wumpus)
Breeze (near pit)
Glitter (gold present)
Bump (hit wall)

Why Wumpus World is Perfect for KBAs:

Partial Observability: Agent must reason about hidden information
Uncertainty: Multiple possible world states
Logical Inference: Can derive safe actions from percepts
Knowledge Accumulation: Each step adds new information

Key Takeaways

Knowledge-Based Agents:

Use knowledge + inference for intelligent behavior
Follow sense-reason-act cycle
Maintain persistent knowledge base
Can learn and adapt over time

Design Principles:

Separate knowledge from control
Use declarative representation
Enable logical inference
Support knowledge updates

Next Steps:

Now that we understand what knowledge-based agents are, we'll explore how to structure and organize the knowledge they use in Topic 2: Knowledge Base Structure.