Computer-Mediated Communication Magazine / Volume 2, Number 2/ February 1, 1995 / Page 3

Talk to My Agent: Software Agents in Virtual Reality

by John Horberg (

As technology develops and attitudes toward virtual spaces change, we will find that our definitions of cyberspace and virtual reality will evolve. But perhaps the most critical element in determining how we create cyberspace and how people see virtual reality lies in the technology for day-to-day use of cyberspace. As we work towards the promise of VR we must continually ask ourselves the questions: What is VR? Where is cyberspace? These are not questions that can be asked or answered easily.

Software agents represent one such piece of technology. Although software agents provide real and useful tools for performing tasks in cyberspace, they are poorly understood by most people. Agent-based technology is making a strong appearance and if we are really interested in cyberspace, we should know something about agents. In addition, software agents can help explicate the nature of VR and its relationship to us and our physical world. Many people have difficulty comprehending exactly what cyberspace is, what it looks like, how we will get there,or and how we will conduct our business there. Software agents offer a unique insight into these areas.

An examination of software agents suggests that VR can be considered a part of our world rather than a separate reality. Cyberpunk fiction and the popular press have created improbable perceptions of the nature of cyberspace; a simpler view based on software agents provides a clearer picture, at least for now. Virtual reality is often defined as a "consensual hallucination" which fools the mind into thinking it is someplace it is not. However, VR might best be seen as a normal state of mind which is working its way into our culture, changing the way we live our everyday lives. Software agents, then, are an ordinary sort of idea that can be useful in facilitating our understanding of cyberspace.

The popularity of cyberpunk fiction has been instrumental in raising public awareness and interest in cyberspace. These stories create expectations of what cyberspace is, how accessible it is, and what uses we will find for it. In addition, cyberpunk fiction has been a driving force behind the development of VR technology, with researchers working to create systems based on the visions of novelists. A reasonable way to proceed is to look at fictional and real accounts of software agents and VR.

What Are Software Agents?

A software agent is simply a computer program that operates while traveling around cyberspace. Individuals, groups, or institutions may own software agents. These owners instruct their agents to roam the networks, access desired information by exchanging data with other agents or people, and handle business and personal transactions. These agents, which are just becoming available on commercial markets, are known by a variety of names: ghosts, intelligent agents, daemons, and elves.

What are the essential qualities of software agents? First, agents are mobile entities. They literally travel around a network and can copy themselves onto other computers to do their work. Second, agents are autonomous entities. They can work on their own, and have the authority to make decisions for their owners. Experts say that we will soon trust software agents with our checkbooks and investment holdings, authorizing them to spend our money for goods and services they know we want to buy. Finally, agents will be intelligent enough to observe us and learn our habits and preferences as well as our personal demographics to serve us better.

These qualities may sound mysterious and rather worrisome, but they are in line with developers' expectations for products that are being created right now. The following sections examine fictional and real world software agents and explain their uses and capabilities.

Software Agents in Fiction

Cyberpunk fiction is filled with descriptions of software agents. Often, though, these descriptions are so exotic that we might not recognize them as agents. Dixie Flatline, the ROM construct in William Gibson's Neuromancer (1984) is one such agent. Dixie perpetually resides in cyberspace and performs jobs for Gibson's "hero," Case. For the most part, Dixie works more quickly and efficiently than Case. Dixie is autonomous and mobile; given a task, Dixie navigates through cyberspace, makes decisions, and reports back to Case when necessary. Like any software program, Dixie is tireless and, more significantly, fallible.

One might argue that since Dixie is actually a recording of an actual human personality "he" cannot be considered a true agent. However, the tasks he performs and the way he interacts with Case make it easy to consider him an example of agent-based technology.

The Librarian in Neal Stephenson's Snow Crash (1992) is another excellent example of software agents in cyberfiction. The Librarian is a single piece of software owned by the character, Hiro Protagonist. The Librarian has a prescribed domain (the Library of Congress) and can access any information in the library to answer a question or to retrieve requested data. Unlike Dixie, the Librarian has a human form in the Metaverse (Stephenson's term for cyberspace), but the interaction between agent and owner takes much the same form: All Hiro has to do is go to his office in the Metaverse, verbally call the Librarian, and tell him what to do. Like Dixie, the Librarian is intelligent enough to speculate when asked to do so, but can not perform tasks outside of his programmed domain.

Finally, Gibson's character, Colin, in Mona Lisa Overdrive (1988) exemplifies the sort of personal assistant that we might expect to become available in the near future. Kumiko, Colin's owner, does not have to enter cyberspace to interact with him. She can simply speak to him (he appears as a sort of hologram) to ask him questions or request that he perform some data retrieval or calculations for her. Colin develops a much more human personality than does Dixie or the Librarian as he learns and remembers much personal information about Kumiko and her environment. His programming requires him to do this as he carries out his task of helping and protecting Kumiko throughout the novel.

Real Software Agents

The above descriptions of fictional software agents represent the kinds of agents that are under development. It is hard to say, at least for now, that real agents will embody the capabilities of fictional ones. However, they operate in much the same way as emerging agent-based technology. Indeed, as voice-recognition technology improves, we might expect to address our agents directly by name as the characters in cyberpunk novels do.

If agent creators succeed, agents will perform such jobs as researching databases, managing investments, buying and selling products, and keeping an eye on almost any specified area of interest.In the future, we will simply place our agents in cyberspace and expect them to report back when something significant happens.

Consider these examples:

Some of these agents are available now or will soon be implemented. Mobile Trader, for example, allows investors to buy and sell investment commodities from almost anywhere, without the aid of a human broker. Investors ask this agent to keep an eye on individual commodities, and to speak up when the time is right to buy or sell (Naik). AT&T's PersonaLink system allows customers to place software agents in a huge data network with instructions to search for products offered at the cheapest prices (Schwartz). Finally, British Telecom is creating corporate and personal agents for its customers. Once developed, a customer can make a phone call, perhaps requesting special services (a video connection or a scrambled line) and their personal agent will confer with the corporate agents to agree on the best mix of services at the best price (Arthur).

In doing their work, your agents will meet with the agents of other individuals and institutions to ask questions, trade data, and buy or sell goods and services. As your agents observe you, they will become accustomed to your habits and preferences. They will do this in order to represent your interests as accurately as possible in the virtual world. Ideally, you agents will act exactly as you would, only with greater accuracy and stamina, to help you cope in a complex world of information.

It is important to remember that, exotic or mundane, software agents are instruments of VR. Agents give us the idea that we are going out into the world and doing something, when in fact we are sitting in front of our computers eating a pizza. That is, agents allow us to go about our business while we wait for them to report back to us. But our questions on the nature of cyberspace revolve around the problems of how we will go about entering and working in cyberspace. Examining some fictional accounts of cyberspace and an agent-based alternative suggests answers to these questions.

Fictional Accounts

Gibson imagines a distinct boundary between the physical world and VR. He constructs this boundary through his terminology for entering and exiting cyberspace: One "jacks in" and "jacks out" of cyberspace, creating an either/or condition. Although Case is able to move from one state to the other just by flipping a switch, his entire consciousness is transported to cyberspace when he is jacked in.

Gibson's cyberspace is also very difficult to work in. Not only do his characters need special equipment and specific skills to enter cyberspace, they must also be highly talented to find and retrieve information. While Gibson paints an exotic and colorful picture of the virtual landscape, he also creates an exclusionary vision of cyberspace, holding VR technology at arm's length from the everyday person.

Stephenson creates a much softer barrier than Gibson does. People need no special skill to enter the Metaverse; its natural architecture means that, for the most part, average citizens can use their normal human skills to get around. While engaged in the Metaverse, all users have to do is peer over their glasses to see what's going on in the real world. Notice, though, that a traveler's consciousness is still totally engaged in Stephenson's "Metaverse." There is still a distinct line drawn between our physical world and a virtual world -- a line that users must cross to engage in activities in cyberspace.

Both of these fictional visions of cyberspace place great demands on the programmers and technologists who must design the architectures of virtual worlds. Not only do these visions require technology capable of fooling the human brain into thinking it is someplace it is not, they demand that cyberspace be navigable through human senses, allow data structures to be located, accessed, and manipulated by real people. Hence the dazzling architecture devised by Gibson. He represents data with geometric forms and lights -- physical structures that can be located and entered. Even Stephenson's more pragmatic view involves a good bit of techno-glitz: million mile-per-hour motorcycles and monorails, slick geometric buildings and fully furnished offices. Both authors create exotic (if desolate) notions of cyberspace that appeal to almost any reader, but seem unattainable with current technology.

An Agent-Based Alternative

The key to the above fictional accounts of cyberspace is that a user's consciousness is transported across a distinct boundary from the real world into cyberspace. Actions performed in cyberspace require the user's attention and real-world time. In keeping with our definition of VR as a consensual hallucination, we like seeing this boundary as concrete, one that we consciously and suddenly step across. Indeed, much of what we think of as VR fosters just that image: VR games require players to don bulky helmets, gloves, and sometimes body suits, allowing them to "step into" a virtual landscape. Surgical simulators for doctors and flight simulators for pilots also have distinct borders: Participants distinctly enter virtual environments and distinctly return to the real world when finished.

However, this perspective defeats the very purpose of cyberspace; if virtual worlds are merely ideal versions of our own world, why bother? The reason for opening up these spaces is to alleviate real world problems. We hope to break the bonds of time and space, making it easier to cope with the huge volumes of information brought about through widespread use of computers and the new information age.

We expect that VR will be available to everyone, not just a few experts. If VR is really going to make its way into the mainstream, it seems likely that the boundary between the physical world and the virtual one will be much softer than either Gibson or Stephenson imagine it -- that cyberspace is a region that we enter gradually and permanently through what becomes everyday technology. As VR products begin to appear and as computers continue to become more deeply entrenched into our lives, we will be drawn slowly, perhaps imperceptibly, into a VR mode of existence.

In fact, we have become quite used to agent-based technology already. Consider a word processor or your telephone answering machine. The word processor places characters and words on a computer screen and into appropriate files as you ask it to. It checks your spelling, formats documents, and sends them to the printer, all on your behalf. Likewise, your answering machine delivers your greeting to callers, records their messages to you, and plays them back whenever you wish.

Another example, the Avis Satellite Guidance system, is a sort of agent working on your behalf to make sure you don't get lost in your Avis rental car. The computer works with a virtual representation of the world, your car's current position, and your destination. All you see, however, is a computer-generated road map and the computer's voice telling you when to turn.

In each of these cases, you do not distinctly enter cyberspace, you simply get an e-mail message or a phone call when an agent has something to tell you. Although it may be difficult to see that these agents voyage into VR, the basic concept is that of cyberspace.

Present Reality

What is VR? Where is cyberspace? In returning to these opening questions we find two broad answers. First, they are exotic and exclusionary spaces that a few power wielding experts will control while ordinary citizens continue to toil in the physical world. Second, VR represents a quiet evolution of the way billions of people go about their daily business.

Considering VR from the perspective of software agents strongly encourages the second, more democratic view. The VR age might be considered in the same way as the telephone age or the automobile age. Certainly the software agents that exist today, as well as those on the drawing board, are leading us closer to daily life with VR.

Computer networks, answering machines, and telecommunications may present a rather drab vision of cyberspace, but they provide us with an exciting vision of how VR is changing our lives. They show that we are beginning, bit by bit, to interact with agent-based technology on a regular basis. As AT&T's "You will" advertising campaign suggests, VR is becoming a real part of our lives aided largely by the development of intelligent software agents. ¤


John Horberg is a PhD student at Rensselaer Polytechnic Institute in Troy, New York. He's a cool guy and a great cook.

Copyright © 1995 by John Horberg. All Rights Reserved.

This Issue / Index / CMC Studies Center / Contact Us