Philosophical agents

Abstraction is the technique we use to deal with complexity. What is the proper kind and level of abstraction for complex software agents? We think it would be reasonable to endow agents with a philosophy. Then, by understanding their philosophies, we can use them more effectively. To endow agents with ethical principles, developers need an architecture that supports explicit goals, principles and capabilities, as well as laws and ways to sanction or punish miscreants. All of the ethical approaches described in this article are single-agent in orientation and encode other agents implicitly     

Consensus software: robustness and social good

In this column I explore some far-reaching issues of software development that lie at the intersection of robust software and sociopolitical systems. These two areas might seem unrelated-and most software developers would likely be horrified to have politics intrude on their programming efforts-but the intersection occurs through these premises: software systems administer and control much of our societal infrastructure; people would appreciate and better accept that control if they had input into the nature of the control and the systems' behavior; designers can make software systems more robust through redundancy, in which different versions of software components might cover for each other's mistakes and limitations; and if many people could contribute software to societal control systems, the systems might be more robust and better represent people's interests. The need for redundancy and the need for widespread participation can be mutually satisfying.     

Commitments among agents

Commitments are a powerful representation for modeling multiagent interactions. Previous approaches have considered the semantics of commitments and how to check compliance with them. However, these approaches do not capture some of the subtleties that arise in real-life applications such as e-commerce, in which contracts and institutions have implicit temporal references. In this column, we describe a rich representation for the temporal content of commitments that lets us capture realistic contracts and avoid ambiguities. Consequently, this approach lets us reason about whether, and at what point, a commitment is satisfied or breached, and whether it is or ever becomes unenforceable.     

Probability and agents

To make sense of the information that agents gather from the Web, they need to reason about it. If the information is precise and correct, they can use engines such as theorem provers to reason logically and derive correct conclusions. Unfortunately, the information is often imprecise and uncertain, which means they will need a probabilistic approach. More than 150 years ago, George Boole presented the logic that bears his name. There is concern that classical logic is not sufficient to model how people do or should reason. Adopting a probabilistic approach in constructing software agents and multiagent systems simplifies some thorny problems and exposes some difficult issues that you might overlook if you used purely logical approaches or (worse!) let procedural matters monopolize design concerns. Assessing the quality of the information received from another agent is a major problem in an agent system. The authors describe Bayesian networks and illustrate how you can use them for information quality assessment     

The Carnot Heterogeneous Database Project: Implemented Applications

The Carnot project was an ambitious research project inheterogeneous databases. It integrated a variety of techniques toaddress a wide range of problems in achieving interoperation inheterogeneous environments. Here we describe some of the majorimplemented applications of this project. These applications concern(a) accessing a legacy scientific database, (b) automating a workflowinvolving legacy systems, (c) cleaning data, and (d) retrievingsemantically appropriate information from structured databases inresponse to text queries. These applications support scientificdecision support, business process management, data integrityenhancement, and analytical decision support, respectively. Theydemonstrate Carnot‘s capabilities for (a) heterogeneous queryprocessing, (b) relaxed transaction and workflow management, (c)knowledge discovery, and (d) heterogeneous resource modelintegration.     

Mobile agents

A lot of agents are executing on the Web, and some of them are starting to move around. While most agents are static (existing as a single process or thread on one host), others can pick up and move their code and data to a new host where they resume executing     

Benevolent agents

Some agents roaming the Web these days are benevolent-for example, they may clean up stalled or failed database transactions, or share query results that may have cost substantial resources to acquire and might consume more to share. The Agent Behavior Testbed is a tool for studying the economics of agent altruism. As more agents hit the Internet, benevolence and cooperation will help with overall efficiency and productivity. The paper discusses benevolent agents on the Web     

A semantic Web services architecture

The semantic Web services initiative architecture (SWSA) committee has created a set of architectural and protocol abstractions that serve as a foundation for semantic Web service technologies. This article summarizes the committee's findings, emphasizing its review of requirements gathered from several different environments. We also identify the scope and potential requirements for a semantic Web services architecture.