You are on page 1of 36

BIS4225.

8
Intelligent Agents

1
What are Intelligent
Agents?
Computational Software
Intelligence Engineering
Intelligent
Agents

Human
Interface
2
Agent Model
(from a user
perspective) Intelligent
Agents

Task Level Communication


Knowledge
Skills Skills
Task A priori Knowledge With User
Learning With Agents

3
Agent Model
Task Level Skills
Information retrieval, filtering, database
queries.
Knowledge Skills
Rules to follow to complete the task.
Communication Skills
Interaction with the user and/or other
agents.
4
Other Descriptions
Environment
The operating environment, e.g.,
operating system, application, network.
Task
What the agent does, e.g., information
retrieval, filtering, search agents.
Architecture
Internal knowledge architecture, e.g.,
learning agents, neural agents.
5
Definition
(No single universally accepted definition.)
A knowledge based system that:
perceives its environment,
reasons (to interpret perceptions),
draws inferences,
solves problems,
determines actions, and
acts upon that environment
to realize a set of goals or tasks.
6
Definition
Two general use of the
term agent:
Weak, relatively uncontentious
(communication with peers by
exchanging messages)
Stronger, potentially more contentious
(implemented using concepts related to
human reasoning)
7
Attributes
Autonomy
can initiate communications,
events, and performs tasks.
Temporal Continuity
Active in foreground or
background.

8
Attributes
Reactivity
respond in a timely fashion to
environmental changes.
Goal-driven
can accept high-level requests
specifying goals of a user, and
decide how to satisfy the request.
9
Classifications
Operating System Agents
Application Agents
Web Search Agents
Information Filtering Agents
Information Retrieval Agents
Notification Agents
10
Classifications
Mobile Agents
Process Automation Agents
Database Agents
Resource Brokering Agents

11
Benefits
Features
Automation
Customization
Notification
Learning
Tutoring
Messaging
12
Business Obstacles
Hyped expectations of users
User Experience
Integration into Business
Models
Security Issues
Privacy Issues
13
Architecture
Intelligent Agent
Input

Environment Inference Engine

Output
Knowledge Base

Challenges
Encoding of knowledge in knowledge base.
Modification of knowledge in response to
changes

14
Learning Agent
Learning Agent
Input
Learning Inference
Environment Engine Engine
Output
Knowledge Base

Learning Agent: An Agent that is able to


acquire and maintain its knowledge base.

15
Learning Agent
Four ways to modify
behaviour:
Monitor user interactions.
Direct and indirect feedback.
Learn from examples given by user
(Training by user).
Check with other agents.
16
Learning Agent
Learning agents operate
on the assumption that:
the application involves a
substantial amount of repetition,
and
the repetition varies considerably
from one user to another.

17
How Agents Work
Simple Agents:
work within the context of a single
application
focus on a single set of tasks with a
circumscribed set of outcomes.
Mainly automates simple repetitive
tasks that could be performed by a
human.
18
How Agents Work
Simple Agents:
The goals of the agent are explicitly
specified by the user.
Applies if-then-else rules.
Actions are invoked by agent
without user intervention.

19
How Agents Work
Rule Base

Event Recognize Pattern Action


& Evaluate Matching

Input Data

20
Autonomy
Agents acting on users behalf
Agents looking for information based
on user criteria
Agents interacting with other services
Agents talking with other agents
As agents makes more decisions,
the higher the risk of errors.

21
Application Areas
Personal Use
Scheduling appointments, locating
information, searching and helper
assistants.
Network Management
Manage network architecture and
traffic.
22
Application Areas
Information and Internet
Access
Manage, filter, select, priortize,
discard, monitor information.
Mobility Management
Manage infrastructure for remote
users.
23
Application Areas
Electronic Commerce
Selecting, comparing prices of
products and services.
User Interfaces
Usable computer interfaces and
shortcuts.

24
Application Areas
Applications Development
Searches for programs, codes,
modules, designs and
specifications.
Project Management
Planning and managing of projects
by offering advice and suggestions.
25
Application Areas
Security and Control
Monitoring
Monitoring devices.
Military Applications
Analysis of sensitive information.

26
Shopping Agents
A2A (Agent-to-Agent
transactions):
Negotiation Protocols
Financial Management
Authentication
Market Mechanisms
27
Designing Intelligent
Agents
Agent Requirements
Analysis: Reduce purpose
of agents to an operational
narrative document about
what it is supposed to do.

28
Designing Intelligent
Agents
Agent Execution
Environmant: There must be
an agent support layer or
agent execution
environment that provides
mobility, verification, and
APIs.
29
Designing Intelligent
Agents
User Acceptance
Agents should:
be unobtrusive
be easily customizable
be useful (progressively complex
tasks)
promote information sharing
30
Agent Standards
Worldwide Web Consortium
(W3C)
ARPA
IETF
FIPA

31
Implementing Agents
Knowledge Acquisition
The process of acquiring
knowledge from a human
expert.
Creation of a model of the
experts domain knowledge.

32
Implementing Agents
Knowledge Representation
The formalization and storing
of knowledge for subsequent
computer processing.
Needs suitable knowledge
representation language to
describe knowledge domains.
33
Implementing Agents
Knowledge Representation
Language:
Representation Adequacy
Inferential Adequacy
Inferential Efficiency
Clear syntax and semantics
Naturalness
34
Implementing Agents
Knowledge Representation
Language examples:
KQML
Telescript/Odyssey
JESS/CLIPS

35
Future of Agents
Areas of Concern:
Future of Network Management
Commercial Agents
Internet searching and data mining
Database agents
Trust
Global Desktop
Agent at home
36

You might also like