Intelligent Agents: Basics & Rationality

1. What is an Agent?

Definition

An Agent is anything that can be viewed as:

1. Perceiving its environment through <abbr title="Devices that detect input like cameras, microphones, keyboards">sensors</abbr>.
2. Acting upon that environment through <abbr title="Devices that perform actions like motors, screens, speakers">effectors/actuators</abbr>.

Examples

Agent Type	Sensors	Actuators	Goal
Human	Eyes, ears, skin, taste	Hands, legs, mouth, body	Survival, comfort, happiness
Robotic Cleaner	Cameras, dirt sensors, cliff sensors	Wheels, brushes, vacuum	Clean floor, avoid falling
Software Bot	Keystrokes, file contents, network packets	Screen display, writing files, sending packets	Filter spam, sort data

---

2. Rational Agents

A Rational Agent is one that does the "right thing".

Definition

For each possible <abbr title="History of everything the agent has perceived">percept sequence</abbr>, a rational agent should select an action that is expected to maximize its performance measure, given the evidence provided by the percept sequence and whatever built-in knowledge the agent has.

Note: Rationality is NOT <abbr title="Knowing everything (impossible in reality)">omniscience</abbr>. Ominscience means knowing the actual outcome of actions. Rationality is about making the best decision with the available information.

Omniscient Agent

An Omniscient Agent knows the actual outcome of its actions and can act accordingly; however, omniscience is impossible in reality. Rationality maximizes expected performance, while omniscience maximizes actual performance.

Example:* An omniscient agent knows everything in advance(including the actual outcome of its actions), so it acts accordingly to always achieve the best result. It creates an impossible standard for real-world agents.

The 4 Factors of Rationality (PEAS)

What is rational at any given time depends on four things:

1. Performance Measure:

* The criteria that determine how successful an agent is.

Example (Vacuum):* +1 point for each clean square, -1 point for each move.

1. Environment:

* Where the agent operates.

Example:* A carpeted room with obstacles.

1. Actuators:

* What the agent can do.

Example:* Move Left, Move Right, Suck dirt.

1. Sensors:

* What the agent can perceive.

Example:* Is the current square dirty? Am I bumping into a wall?

Ideal Rational Agent

An ideal rational agent always chooses the action that maximizes its expected performance, based on what it knows and sees.

---

3. Structure of Intelligent Agents

An agent consists of two main parts:

Formula

Agent = Architecture + Program

1. Agent Program:

* The internal logic or algorithm.

* Function: Action = AgentFunction(Percept)

* It implements the mapping from percepts to actions.

1. Architecture:

* The physical computing device (hardware).

* Examples: Computer, Robot body, Server.

* The architecture makes the percepts available to the program and runs the program.

---