Software Agents

Overview

Agent as a black box


Intelligent Agents

Human Agents (e.g. travel agent)
Hardware Agents (e.g. robot)
Software Agents

Information Agent
Cooperation Agent
Transaction Agent

Information Agent:

support in search for information in networks.

Cooperation Agent:

complex problem solving with other agents or humans

Transaction Agent:

processing and monitoring of transactions

Characteristics of Intelligent Software Agents

Reactivity (senses environment and reacts)

True reactive (no internal environment model)
Deliberative (internal environment model)

example: Pointcast Network

Proactivity/goal-orientation

Reasoning/learning

Autonomy (follow goals by acting alone)

Mobility (move from one computer to another over a network)

Communication/cooperation


Deliberative agents

Beliefs (about environment)

Desires (judgment of future situations, may be unrealistic or contradictory)

Goals (subset of desires on which agent could act)

Intentions (subset of goals chosen to act upon)


Related areas

Artificial Intelligence (learning, reactivity, proactivity)

Distributed Artificial Intelligence (communication, cooperation)

Network comunication (communication, mobility)

Decision theory (Autonomy, learning)

back to Overview


Agent Examples

Simple Search engines

Input of information: follow links
Indexing of information: syntactic analysis, storage in database
Retrieval: compare query with doc contents, rank retrieved documents
www.altavista.com , www.webcrawler.com , www.hotbot.com

Meta Search engines

Query multiple simple search engines

Aggregate results (relevance check, recognition of duplicates)

www.go2net.com (MetaCrawler)

NewsWatcher

Personalized messages and information items on the Internet

www.pointcast.com

Recommender or Collaborative Filtering Agents

Finding movies that you like

User interface agents

Letizia

back to Overview

E-commerce Agents

Shopping agents for comparison shopping

Jango Junglee

Agent-based marketplaces, negotiation, collaboration, auctions

MarketMaker, Tete-a-Tete

back to Overview

Mobile Agents

Agents that transport themselves over the Internet.

Mobile network agents are programs that can be dispatched from one computer and transported to a remote computer for execution.
Arriving at the remote computer, they present their credentials and obtain
access to local services and data.
The remote computer may also serve as a
broker by bringing together agents with similar interests and compatible goals, thus providing a meeting place at which agents can interact.
(from
IBM Aglets)

back to Overview

A government case study: FinCEN, a system for finding financial crime.

Cash transactions over $10,000 are reported and stored in database

Database size is approx. 10M

Suspiciousness evaluation rules

Both data-driven and user-directed

GIS and link analysis & visualization tools

Key problem: Group transactions into related sets for detailed analysis

Consolidation: Matching of fields to identify all transactions involving people, businesses or accounts

Linkage: Identification of pairs of subjects and/or accounts which share a common transaction

Data mining done by humans

Future directions:

Use an agent-based approach for consolidation and linkage.

Use FinCEN's results to update the rule base.

 

References

Intelligent Software Agents Application and Classification Nick Moraitakis

Agent Tutorial by Pattie Maes

Software Agents for Information Retrieval Tutorial by Tim Finin, C. Nicholas, J. Mayfield