Transcript Document
E-Learning
Intelligent Agents and E-Business
Intelligent Agents
What are they?
How do they work?
Why e-business?
Assumptions
Summary
What is an agent?
Human agent or assistant
travel agent
estate agent
secret agent
Focus on a task
Specialist skills
Access to relevant information
Contacts to provide service, can provide
it at fraction of cost
Can only work < 84 hours per week
What can an agent do?
Provide information & description of a
service or product
Locate best sources, companies,
locations
Find best prices
Negotiate agreement between
purchaser and supplier
Prepare & distribute documents,
contracts, agreements
What can an agent do
cont.
Monitor results and resolve problems
Provide additional information and
clarification
Collect revenue, fees, commissions &
distribute funds
Terminate service/product if no payment
Send out renewals & reminders
Example
Solicitors
House Seller
House Buyer
Buying
Agent
Selling
Agent
What are intelligent
agents?
software or hardware applications
undertake tasks on behalf of user
possess some knowledge
may incorporate a learning system
use reasoning to achieve user’s goals
Simple example:
email filter - directs emails into folders
automatically based on certain pre-defined
criteria
Intelligent agents cont.
Agents typically act independently of
each other
learn and adapt from actions or the
environment
can be mobile
Categorised using three dimensions:
Agency
Intelligence
Mobility
Agency
‘Degree of independence’
knowing the user’s goals, must be able to
accomplish a mission without human
intervention
need to communicate with data
repositories and other agents
increased sophistication allows
cooperation between agents, sharing
knowledge and objectives to solve
common goals
Intelligence
Ability to reason using current
knowledge of user as well as past
experiences
Could be rule-based;
reasoned against a set of pre-defined
conditions
Knowledge-based;
data relating to prior situations and
resulting actions supplied to agent, on
which it bases its decisions
Mobility
Agents can be static, working in one
computer
Can also travel across networks,
gathering information
Some travel from computer to computer,
returning to host only when all search
conditions are met
Others travel between client and server
(more secure)
Other Attributes
Act in the best interests of the user
must not injure a human, or through
inaction allow a human to come to harm
must obey orders given by humans except
where such orders conflict with the first law
must protect its own existence as long as
such protection does not conflict with the
1st and 2nd laws
How do they know what
to do?
User programmed approach
needs to represent own knowledge in a
programming language
Knowledge-based approach
provide agent with user and domain (environment)
knowledge
difficult to customise - generic solution
Learning approach
minimum background information
learns from user and other agents
behaviour must be repetitive
Recap
1. PA not familiar with
habits/systems of new
employer - not very
helpful on day one
2. PA needs time to
become familiar with
working methods
1. Agent is ignorant
2. Agent needs either
pre-defined knowledge
base or time to learn
4. More tasks performed
by employer are taken
over by the PA
3. PA Learns by watching
employer perform tasks,
receiving
instructions/training and
learning from other PAs
4. Agent adapts and
becomes more
autonomous
3. Agent interacts with
user and other agents
How do they work?
Develop a user profile
where you visit
what you read/watch/contribute
Learn
how user evolves
how similar users evolve
How do they
communicate?
Need a common transport protocol
messages sent by one party must be
understood by others
TCP/IP
Can use HTML to exchange information
How do they work?
Client allowed
to store
procedures,
instructions
and data on
server
Client
accesses
agent via
browser on
workstation
HTML message sent via TCP/IP
User profile stored on client or server - depending on
system architecture
Data on client - less network traffic, but data could be
lost
Data on server - more traffic, data backed-up, but
DPA and hacking/misuse issues
What have they got to
do with e-business?
Buyer agents
personal shopper
reduced effort - ‘Let your fingers do the walking’,
Yellow Pages
comparing prices while you sleep
Seller agents
target your market more effectively
dynamic market research, knowledge mining
learn from customers and personalise
Assumptions
Design
need to incorporate open design to allow
continuous development
Quality of Databases
quality of content across the web is
extremely variable as are standards
sites need some authentication for
trustworthiness
need consistent data standards - XML
Assumptions 2
Functionality
optimum benfits derived when intelligence
gathering is incorporated
Friendly environment
for multi-agent environments, hosts must
allow agent scripts to live and work
remotely
must converse with visiting agents
web security a major issue
Summary
What are they?
human and software agents
How do they work?
agency, mobility, system architecture,
profiles
Why e-business?
buying, selling, market research
Assumptions
standards, databases, remote working,
security