Middle-Agents for the Internet.

Like middle-men in physical commerce, middleagents support the flow of information in electronic commerce, assisting in locating and connecting the ultimate information provider with the ultimate information requester. Many different types of middleagents will be useful in realistic, large, distributed, open multi-agent problem solving systems. These include matchmakers or yellow page agents that process advertisements, blackboard agents that collect requests, and brokers that process both. The behaviors of each type of middle-agent have certain performance characteristics—privacy, robustness, and adaptiveness qualities—that are related to characteristics of the external environment and of the agents themselves. For example, while brokered systems are more vulnerable to certain failures, they are also able to cope more quickly with a rapidly fluctuating agent workforce and meet certain privacy considerations. This paper identifies a spectrum of middle-agents, characterizes the behavior of three different types, and reports on initial experiments that focus on evaluating performance tradeoffs between matchmaking and brokering middle-agents, according to criteria such as load balancing, robustness, dynamic preferences or capabilities, and privacy. 1