Software Agents
Agenda
Agent Visionaries
Video Knowledge Navigator
What is an Agent?
Common Issues Studied
Types of Agents
What is a Software Agent?
How is an Agent different from other Software?
Why do we need Software Agents?
Direct Manipulation
Indirect Management/Agents
Criticisms of Software Agents (Lanier, Schneiderman)
Software Agent =? Expert System
Agents vs Expert Systems (cont.)
Types of Software Agents
Types of Software Agents (cont.)
User-Programmed Agents
Nature of “Intelligence” (cont.)
Knowledge-Based Agents
Learning from the User
Learning from other Agents
Example: Email Agent
Second Example: News Agent
Real example of User-Programmed Agent: OVAL (Malone, MIT Sloan)
OVAL (cont.)
PPT Slide
Other examples of User Programmed Agents
Real Example of Knowledge-BasedAgent: CHORIS (Tyler & Sullivan, Lockheed)
Other examples of Knowledge Based Agents
Real Examples of Learning Agent: Maxims (MIT Media Lab)
Maxims Learning Agent (cont)
Learning from Peers: Peer-Peer model
Other examples of Learning Agents
Pros/Cons of Approaches
Which approach is best?
Location of Agents
Location of Agents (cont.)
Mobility of Agents
Mobility of Agents (cont.)
Exercise
Technical challenges
Software Agent <-> Application
Technical Challenges
Interface Agent <-> Other Agent
Common Language, Common Ontology
Negotiation & Commitment methods
Modeling other Agents
Authentication & Identification
Interface Agent <-> User
Issue 1: Understanding
Issue 2: Control
Issue 3: Distraction
Issue 4: Ease of Use
Issue 5: Personification
Video: Persona Project (Microsoft)
Personified Agents Research
Roles for Software Agents
Agents as Eager Assistants
Eager (Cypher)
Video Eager
Video Meeting Scheduler (MIT Media Lab)
Maxims Email Agent (MIT Media Lab)
Eager Assistant Agent (cont.)
Details of one Example
Prediction & Confidence Level Computation
Using the Prediction
Multi-agent Collaboration
Agent Interface
Explanation
Video Maxims
Does it work? Results:
Discussion (cont.)
Video Schlimmer’s Note Taking Agents
Eager Assistants: Dimensions
Agents as Guides
Video Guides (Apple)
Letizia (Lieberman)
Video Letizia (Lieberman)
Feature-Based Filtering: Analyzing Documents for Relevant Keywords
Feature-Based Filtering: Updating User Profile Based on one More Datapoint Document
Feature-Based Filtering: Filtering Based on User Profile
Agents as Memory Aids
Remembrance Agent (Rhodes/Starner)
Remembrance Agent URL
MIT Media Lab Wearables Group
Agents as Filters/Critics
Video NewsTaylor
Technology Underlying Critics: Feature-based Filtering (FBF)
Demo Firefly and Webhound
Complementary Technology Underlying Critics: Automated Collaborative Filtering (ACF)
ACF - Predicting a Rating
Collaborative vs Feature based
Comparison (cont.)
Some other Neat Features
Neat Features (cont.)
Other quantitative Results
Quantitative Results – Base case Algorithm:
Quantitative Results – HOMR/Ringo:
Agents as Matchmakers
Yenta (Foner)
Yenta applications
Yenta Methods
Yenta Algorithm
Agents for Buying/Selling
Bargain Finder demo
Fido shopping Doggie Demo
Kasbah (Chavez)
Kasbah Example
The Future of Agent Technology
Future Direction: Markets for Agents
Future Direction: Ecologies of Agents
Implications of Software Agents
Benefits to Consumer
Benefits to Marketers/Producers
Speculations: Effects of Agents on Markets
Open Questions
Further References
Email: pattie@media.mit.edu
Home Page: http://www.media.mit.edu/~pattie
Other information: Software Agents Group MIT Media Laboratory