Agents on the Web: Trust and Persistence

W e rely on computers to control our power plants and water supplies, our automobiles and transportation systems, and soon our economic and political systems. Increasingly , software agents are enmeshed in these systems, serving as the glue that connects distributed components. They translate messages, manage database queries and updates, represent buyers and sellers in e-commerce transactions, and may even cast votes on our behalf. Clearly, we need mechanisms to determine whether these agents are trustworthy. What do we need to establish trust? Agents are often characterized by features such as autonomy, sociability, proactiveness, and persistent identity. This latter feature is key in determining trust. When agents operate over an extended period, they can earn a reputation for competence, timeliness, ease of use, and trustworthiness—something ephemeral agents cannot do. Along with persistence, we need a reliable way to identify an agent and ensure that its true identity is not concealed. Researchers have addressed this issue in the context of mobile code, but only to the extent of protecting against malicious code alteration or behavior. How can we assess an agent's trustworthiness? As with other aspects of agents and multiagent systems , we can take our cue from the human domain. Our reputations for trustworthiness are determined and maintained by the people we deal with. Analogously , a software agent's reputation will reside within other agents with whom it interacts. Obviously , agent reputations can't exist without persistence and unique identities. The importance of reputation should not be underestimated: for both humans and agents, a good reputation serves as a social lubricant. As agent research and technology advance, we will ask agents to solve increasingly complex problems. An agent deemed trustworthy will find it easier to establish relationships with other agents. These relationships provide access to resources that can assist an agent with its task at hand, just as a friendly relationship among neighbors can result in mutual assistance. For some agent interactions, such as those involving commerce, agents will simply inherit the reputation of their human owner, sharing, for example, their owner's credit rating and financial capability. For other types of interactions, such as those involving information gathering, an agent will determine its own reputation through its efforts at gathering and distilling information. An agent with a reputation for conducting thorough searches will be trusted by other agents wishing to use its Web search results. Envisioning Agent Societies Multiagent systems …