intelligent agents: an overview

29
Intelligent Agents: Intelligent Agents: An Overview An Overview From: Chapter 1, A. Canlayan and C. Harrison, Agent: Sourcebook, Wiley 1997.

Upload: boris-dean

Post on 02-Jan-2016

50 views

Category:

Documents


0 download

DESCRIPTION

Intelligent Agents: An Overview. From: Chapter 1, A. Canlayan and C. Harrison, Agent: Sourcebook , Wiley 1997. Contents. Attributes of intelligent agents End user taxonomy of agents Intelligent agent applications Benefits of agents Business obstacles for agent acceptance - PowerPoint PPT Presentation

TRANSCRIPT

Intelligent Agents:Intelligent Agents:An OverviewAn Overview

From: Chapter 1, A. Canlayan and C. Harrison, Agent: Sourcebook, Wiley 1997.

2

ContentsContents

• Attributes of intelligent agents• End user taxonomy of agents• Intelligent agent applications• Benefits of agents• Business obstacles for agent

acceptance• Agent use prediction• Cost of development

3

BackgroundBackground

• A definition of an agent:

Agent: A person or thing that acts or is capable of acting or is empowered to act for another.

The Webster’s New World Dictionary 1970

• Two key attributes are pointed out:– An agent does things.– An agent acts on behalf of someone or

something.

4

BackgroundBackground

• Intelligent agent that resides on computers always incorporate these two central attributes.

• The following definition of an agent will suffice to discuss the business applications of an agent:

Software agent: A computing entity that performsuser delegated tasks autonomously.

• Mail filtering agents, information retrieval agents, and desktop automation agents all fit this definition.

5

Attributes of Intelligent Attributes of Intelligent AgentsAgents

• The agent possesses the following minimal characteristics:– Delegation– Communication skills– Autonomy– Monitoring– Actuation– Intelligence

6

Attributes of Intelligent Attributes of Intelligent AgentsAgents

• The concept of an agent introduces– an indirect management metaphor in a c

omputerized environment– to supplement today‘s mainstream styl

e of direct manipulatioon metaphor via GUI.

(Alan, K. (1984). “Computer Software,” Scientific American (March).)

7

Attributes of Intelligent Attributes of Intelligent AgentsAgents

• The origins of agent technology are rooted in– the computational intelligence,– software engineering, and – human interface domain.

8

IntelligentAgents

NeuralNetworks

KBSIntentional

Systems

ReasoningTheory

ComputationalIntelligence

SoftwareEngineering

HumanInterface

Objects

Image a

nd

Speech

Proce

ssin

g

High-level

Event

Inferencingon-lineMonitoring

IntelligentTutoring

Inte

ract

ive

Experim

entsCognitive

EngineeringUserModeling

9

Attributes of Intelligent Attributes of Intelligent AgentsAgents

• Agent model from a user perspective

10

Agent

Task LevelSkills Knowledge Communication

Skills

TaskA Priori

KnowledgeLearning With User

With OtherAgents

Information RetrievalInformation FilteringCoaching

Developer SpecifiedUser SpecifiedSystem Specified

Dialog BasedMemory BasedNeural NetworkCase-BasedNeural Expert

InterfaceSpeechSocial

InteragentCommunicationLanguage

11

End User Taxonomy of End User Taxonomy of AgentsAgents

• It is helpful to define the agent environment, task and arachitecture.

• Environment: Agents are designed to performed in a particular environment such as an OS, an application, and a computer network.– Internet agets, OS agents, WWW agents– Assistants, experts, and wizards for a given appli

cation

12

End User Taxonomy of End User Taxonomy of AgentsAgents

• Task: Task-specific agents are named accoding to what the agent does.– Information filtering,– informaton retrieval, and – search agents

13

End User Taxonomy of End User Taxonomy of AgentsAgents

• Architecture: Agents are labeled according to the internal knowledge architecture.– Learning agents– Neural agents

14

End User Taxonomy of End User Taxonomy of AgentsAgents

• Taxonomy in this book:– Desktop agents:

• OS agents: interface agents that provide user assistance with the desktop OS

• Application agents: interface agents that provide assistance to the user in a particular application

• Application suite agents: interface agents that help users in dealing with a suite of applications

15

End User Taxonomy of End User Taxonomy of AgentsAgents

– Internet agents:• Web search agents• Web server agents: Internet agents that reside at a spe

cific Web site to provide agent services• Information filtering agents• Information retrieval agents• Notification agents• Service agents• Mobile agents: agents that travel from one place to an

other to execute user-specified tasks

16

End User Taxonomy of End User Taxonomy of AgentsAgents

– Intranet agents:• Collaborative customization agents: intranet agents t

hat automate workflow processes in business units• Process automation agents: intranet agents that atom

ate business workflow processes• Database agents: intranet agents that provide agent s

ervices for users of enterprise databases• Resource brokering agents: agents that perform resou

rce allocation in client/server architecture

17

Benefits of AgentsBenefits of Agents

Feature Advantage Benefit

Automation Perform repetitive tasks I ncreasedproductivity

Customization Customize informationinteraction

Reduced overloaded

Notification Notif y user of events ofsignificance

Reduced workload

Learning Learn user(s) behavior Proactive assistance

Tutoring Coach user in context Reduced training

Messaging Perform task remotely Off -line work

18

Benefits of AgentsBenefits of Agents

• Automation– Particularly applicable for automating:

• Repetitive behavor of single user• Similar behavior of a group of users• Repetitive sequential behavior of a number of users in

a workflow thread– Repetieive behavior can be

• time-based or• Evenet-based

19

Benefits of AgentsBenefits of Agents

• Cutomization– Fit into the traditional broadcast and publishing

models.– There are three basic architecture choices in the

implementation of such a model:• the agents can be implemented at the broadcast site,• at the user end, or • in the middle as a broker agent that serves multiple br

oadcaster and users.

20

Benefits of AgentsBenefits of Agents

• Notification:– For instance, such an agent can monitor evetnts

of personal changes, and report them to a user.

21

Benefits of AgentsBenefits of Agents

• Learning:– An agent with a learning capability can learn tasks that can

be automated or preference that can be used for customization:

• Learning and offering to automate the repetitive tasks of a single user, this releiving the user of the need to toil with what, when, and how to automate

• Leanring the similar attributes of a group of users to customize information based on group characteristics

• Learning similar behavior of a group of users to provide workgroup productivity enhancement

• Learning and offering to automate recurrent sequential behavior of a group of users in a workflow thread, thus relieving the workgroup of repetitive tasks

22

Benefits of AgentsBenefits of Agents

• Tutoring– An agent with a tutoring capability can coach a

user in context thanks to its event monitoring and inferencing capabilities, thus reducing the training requirements.

– For example, application wizards in the Windows OS

23

Benefits of AgentsBenefits of Agents

• Messaging– A messaging agent enables user to accomplish t

asks off-line at remote sites.– Mobile agents are examples of messaging agent

s that can transport themselves from place to place to interact with other agents to perfrom tasks on behalf of a user.

24

Business Obstacles for Agent AccBusiness Obstacles for Agent Acceptanceeptance

• Hype– The concept of an intelligent is easily grasped by anyone,

and generalized freely.– Users do not care about the complexity in being able to d

eliver such functionality across all applications.– Unfortunately, the delived functionality cannot easily kee

p up with the generalized expectations of users.– The solution is to focus on task-specific agents for narr

ow domains.

25

Business Obstacles for Agent AccBusiness Obstacles for Agent Acceptanceeptance

• User Experience — Indirect Manipulation– A new human-computer interaction beyond today‘s dire

ct manipulation metaphor with GUI– Mass market acceptance of a change in user experience d

oes usually take a number of years.• Business Model• Security• Privacy

26

Agent Use PredictionsAgent Use Predictions

• Software agetns will be accepted as a design paragigm like object-oriented programming or client/server computing according the following observation:– Task-centered computing is slowly replacing the current

application centered computing paradigm.– The move toward document-centered computing with OL

E and HTML will accelerate this trend.– The software agent model is a better fit to task-centered c

omputing than the current application software model.

27

Agent Use PredictionsAgent Use Predictions

• Prediction for the desktop:– SA will be incorporated into task-specific applications to

provide apllication-specific assistance.– SA will supplement today‘s GUIs with intelligent backen

d services, for example, MS wizards.– This replacement will be very much like the replacement

of command line ionterface software with applications supporting industry-standard GUIs.

28

Agent Use PredictionsAgent Use Predictions

• Predication for Intranet– Agents will emerge as critical components of wo

rkflow solutions within the enterprise.– Task-specific agents will serve as intelligent fron

t ends to enterprise information systems.– Internet-based agents will get modified for intra

net applications to manage the specialized information needs of the corporation.

29

Agent Use PredictionsAgent Use Predictions

• Predication on the Internet– Agents, in the short term, will emerge as inform

ation brokers for specialized domains implemented as centralized Web services.

– In essence, agents will be components of Web-based services incorporating agent functionality.

– Web search engine exemplify such a trend.