The software agent manifests in many forms and aspects, but its main purpose is to undertake action(s) on behalf of a person(s), without having to be told what to do or how. Certainly there will always be some initial setting of user preferences, and the possibility of modifications later, but the point is a person can start up an agent, point it in a general or specific direction, and let it go about its work. The difference between agents and typical computer software is that some agents never need direct and deliberate instructions from users, but rather are designed to watch and learn user behavior and then act accordingly. The end result of agent mediated computer use is that the computer experience will become more tailored to users, at least to the extent that agents are capable of modifying and interacting on their user's behalf.
But what exactly do agents do? Well, like expert systems, there are various areas of expertise, domains of influence, purposes and motivations with which they are programmed. While the actual use of agents is rather limited at the present to companies that are aware of the technology, there is a vast array of agents that are being designed and tested. One of the foremost experts and researchers in this field is Dr. Patricia Maes who works at the MIT Media Lab specifically in the Software Agents Group. While there are all kinds of projects going on in the Software Agents Group, one of the main ideas that Maes is working on is "community ware", an invisioned "electronic knowledge marketplace." Maes sees the uses of agents reaching the point where individual users will have their own knowledge agents that wil actually move about on the Internet and interaction with other knowledge agents in the hopes of procuring the knowledge its user instructed it to find. Thus, professors and other such experts could send out knowledge agents complete with field of expertise, price for consultation, hours available for consultation, etc., and then non-experts could send out their own agents set to find particular knowledge, price willing to pay for the knowledge, etc. The "selling" agents and the "buying" agents would interact and come to a mutual agreement on behalf of their users, and then a knowledge transaction could take place. Instead of putting ads in newspapers, or even surfing the 'net oneself, agents can be activated with a purpose or goal and do the looking, buying/selling, and logistical arrangements themselves! In fact, Dr. Maes and others have even started a business selling agent software to other companies, but even if you aren't a business, there is also some agent technology available to the common web denizen at Firefly Network, Inc.
While Maes' knowledge marketplace is
just one example of the potentials of this new technology, there are other less
sophisticated examples of software agents that aren't so uncommon or futuristic. In
fact, if you have used the web to any great extent, it is most likely that you have taken
advantage of agent technology. Bots, a shortened form of robots, are one of
the more common place agents on the web, though some can be found lurking in various MUDs
(Multi-User Dungeons). Check out the Botspot
for the latest on bots. At the Botspot you can find links to various types of
A general description of each type of bot is also available.
Intelligent Objects are the next real world applications, after exper systems, and they benefit from all that has gone before, but especially software agent technology.
This project was produced for PSY 380, Social Psychology of Cyberspace, Spring, 1998, at Miami University.