One alternative considered for the Kasbah trial was to not actually give participants the items they traded but instead to tell them what they owned, keep track of their inventory for them, and hand out the items at the end of the day. On the face of it this approach has some appeal: it saves people the trouble of carrying around bulky objects. A similar question was considered for the money in the system: should we give out physical bills to each participant or simply keep track of balances electronically? While keeping the objects of trade virtual would not change the functional aspects of the marketplace it would have had a large effect on the psychological impact, a complete change in the relationship people had with the items they were trading.
In the end the decision was made to choose the physical route, giving out real items and paper currency. This choice seems to have been popular, at least with respect to the items. Participants spent a fair amount of time with the items they owned, flipping through the books, using the cameras and bags. People responded to actually having the objects, they had a sense of direct ownership that would have been difficult to achieve if they were only going to actually receive the items later in the day.
The success of the physical cash was less clear. Participants were not often seen looking through their wallets. The actual physical transference of cash was awkward. A peculiar aspect of the Kasbah system was that because the currency was play money, it not only had no intrinsic value but at the end of the day had no consensus value either. The cash was only good for the duration of the trial. As expected over the course of the day the prices of the items inflated. Participants indicated that the physical cash had little meaning to them. Perhaps this is simply because people are already used to the abstraction of money (via experience with credit cards). If Kasbah had been run with electronic books it likely would have had little impact on people's experiences.
An important goal of Kasbah was to get participants to meet each other and socialize; it was hoped that the trade of objects would facilitate random encounters. The software agents met at random to negotiate a trade. The intention was the agents would notify their owners, who would then meet in the physical world and perform the exchange. The objects of trade would serve as introducers. In practice physical meetings proved awkward. Most people chose to take advantage of the centralized bank that would complete halves of the trade; even when the two people in a particular trade were standing next to each other they often used the bank to perform the transaction. In retrospect it seems likely that the difficulty of finding a person and the lack of incentive to trade in person combined to discourage chance meetings. The physicality of the tradable objects was useful in that it made the Kasbah trial more engaging for participants but other barriers kept the objects from fulfilling the desired role of an introducer of people.
The significance of the physical space of the Kasbah trial cannot be overestimated. The impact of people being in the same room affected the entire experience in profound ways. Given that the eventual deployment of Kasbah is intended to be Internet-wide, without physical space, it is important to understand what effects the physical environment had on the use of the system.
Because Kasbah was only one event competing for attention during the day it was important to distinguish participation in the marketplace from other aspects of the symposium. The physicality of the environment neatly did this by dividing the space up into the general lobby area for the symposium and the particular place for Kasbah itself (the Cube). These two physical spaces related directly to the activities of participants. People deliberately walked into the Cube when they wanted to be in the marketplace, changing their physical environment in combination with changing their attention. The ambiguous space between the lobby and the Cube served as a useful intermediary environment, a place where people could participate partly in Kasbah but also stay in the more general milieu of the symposium.
The most obvious aspect of the shared room is that it created a social experience. The Cube was a place where participants were in close quarters, could watch each other, talk about what was going on, and participate in the general buzz of the market that made the event compelling. This sort of community experience depends on having a space, a physical marketplace. Creating a simulation of this experience in a virtual world could be possible but it comes for free in a physical world.
Explicit interface objects are a peculiarity of virtual worlds; in a traditional marketplace the objects of the market are themselves the interface. In Kasbah, the interface objects serve as intermediaries between the virtual marketplace inside the computer and the real-world experience of the participants in the market. In designing Kasbah the creation of the interfaces was the hardest part. The lessons learned from observing the participants interactions with the interface objects are helpful in considering how to build effective virtual marketplaces.
The primary interface to Kasbah was a group of kiosks, computers running Web browsers. Participants walked up to a kiosk, scanned their badge with a barcode reader, and proceeded to manipulate a series of web pages that were their primary control interface to the market and their software agents. The intention was to make the kiosks as simple as possible. Only a mouse and a barcode reader were needed for input and the output web pages were carefully designed to be straightforward. In practice participants were able to use the computers with little or no instruction. Given how easily technologically literate people interact with computers it is important to remember how specialized this skill is. Participants had to understand abstract metaphors of action (clicking on text), display (web pages, chunks of information), and presence (logging into the server for their own personal information). The fact that all of these metaphors were accepted without difficulty indicates how quickly our society (or at least, the subset represented in the trial) has adapted to the new paradigms of computer interfaces.
In order for people to be visible to the computer kiosk they had to somehow tell the computer who they are. The traditional way to do this is to type in a username, but this is difficult if you have no keyboard. The solution chosen by the Kasbah team was to give participants name-tags that included barcode representations of their names. For a person to log in, he or she simply walked up to the computer and scanned the tag. Before the trial there was worry that the barcodes were vaguely creepy: barcodes are impersonal, alienating, imply labelling people like products are labeled in stores. Why should people go out of our way to be visible to computers? But during the trial participants did not seem to even notice the barcodes, were happy with the convenience of scanning their badges. Maybe participants were particularly willing to experiment with technology in the context of a visit to the Media Lab, or perhaps barcodes have been accepted in common culture.
In order to give participants a feel for how the market as a whole was behaving two ``big board'' displays, large computer screens, were deployed in the space. The displays alternated between two kinds of information: a time series of the price of each item over time and histograms showing the current bids and asking prices for each item. The alternation speed was chosen to be slow enough so that the information was comfortably readable but fast enough that it felt like the displays were displaying current information. The primary goal of these displays was to try to summarize the market effectively, give people a sense for the dynamics. As an abstract information display it was a success.
But the physical nature of the displays proved to have important functions. The presence of the monitors themselves focussed attention, gave people a place to direct their gaze when wanting to observe the market and think about how it was performing. Participants were regularly seen standing around intentionally looking at the screen: the display served as a physical manifestation of the abstract virtual marketplace. In addition the displays were social focus points. Participants regularly stood in clusters in front of the screens, pointing and talking to each other about their theories. The displays were therefore useful in facilitating social contacts, more so than the trade objects themselves.
Another market display similar to the big boards were scrolling tickertapes that displayed the current bid and asking price for each item. These displays were a conscious imitation of the archetypical stock market tickertape displays seen in stock exchanges and on the bottom of television programs like CNN Headline News. There were two kinds of tickertapes: Java applets on the web pages at the kiosks and two LED boards in the physical space near the big board monitors. The actual data displayed on the tickertapes did not seem to be used by many participants as there were more convenient ways to get the same information. But the presence of the tickertapes served a useful iconic function. They made a visual statement that Kasbah was a marketplace like other markets people are familiar with. At one point during the day the tickertape display broke so that the text being scrolled was an error message instead of market data. While this was mildly embarrassing it did not seem to disturb participants. At worst they were amused at the clash of contexts between tickertape and computer error messages.
A final important interface were alphanumeric pagers given out to Kasbah participants. The pagers were the way agents proactively communicated with the people they were working for. When a transaction was made or an agent decided it was not going to be able to easily find a deal it would send a message out causing the pager in the participant's pocket to vibrate and display a text message. Pagers proved to be fairly useful interface technology: they were unobtrusive and small although the reading of messages themselves on pager was awkward. The pager objects served as extensions of the agents into the physical world, partially embodying the agents.
Most of the components going into Kasbah, the trade items, physical space, and interfaces, were straightforward and familiar enough to not present many problems to the users. By contrast the trading agents themselves were new to most of the participants, disconnected and unfamiliar. It was important for the Kasbah trial that the idea of the software agents was communicated as simply as possible. In retrospect it was easy enough to explain the functional role of agents but their ambiguous virtual status made people's interactions with them less than transparent.
The most direct way that people contacted with their agents was via the web interface on the kiosks. The physical characteristics of the kiosks have already been discussed; they were effective if prosaic. However the same familiarity of computer interface that made the kiosks easy to use possibly hampered the interaction between people and their agents. Trading agents are fundamentally dynamic entities: they possess goals, strategies, and behaviors. But the representation of the agents on the kiosks was static, showing only the barest configuration details of each agent. The gulf between static representation and dynamic behavior seems to have made communication between people and their agents difficult.
An added confusion was that while the metaphor of agents is of people browsing a marketplace to make a deal the actualization of that metaphor was obscure. The agents did not seem to be located anywhere, certainly not in a familiar physical space. In a sense the agents were located in the virtual space of the computer marketplace, but the virtual marketplace was too abstract and did not seem to be conveyed well. Participants could see their agents at the kiosks, but the representation there was of the static agent isolated from the virtual interaction space. Looking at the big board display conveyed some information about how the agents interacted but the connection was difficult for people to understand. The location of the trading agents was problematic.
Kasbah agents are intended to be autonomous -- a person creates an agent with all the necessary instruction and then sets it free. This autonomy is at odds with tinkering with the agent marketplace Autonomous agents deliberately discourage full hands-on experimentation with the structure of the agents themselves. In practice, most participants in the Kasbah trial had other things to do and did not have the time to play with the market. A few people did make the time to experiment, changing their existing agents and creating new ones to try to find the best deal or manipulate the market price. These people seemed to be somewhat satisfied with their ability to modify the agents but disappointed in the lack of tools and access to really get inside and modify the agent behavior.
Software agents research frequently talks about personifying agents, endowing them with some of the psychological status of people as a way to define a metaphor for their action. The Kasbah agents were quite trivial, acting only in the limited domain of transaction negotiation. As such the status of their personification difficult, more of a cartoon view of what it means to be a person. During the Kasbah trial no one seemed to think of their agents as metaphorical people and the various personifications built into the system were ineffective. It is telling that even though every communication from an agent contained its name, participants never referred to their agents by name. One participant remarked that he ``just wanted to know how his agent felt,'' trying to understand the current status of an agent via an emotional metaphor. The Kasbah system did not make that personification easy.
The the Kasbah trial was successful in part because it deliberately tried to evoke traditional real-world markets. Giving users real objects to trade and having them together in a physical room were successful strategies: these aspects of Kasbah were exactly like a physical marketplace. Underlying the physical Kasbah was the virtual agent marketplace. The connection between the virtual and physical was important to make the event a success. The interface objects in the system were effective in bringing the virtual marketplace into the familiar physical world, particularly the big board display and the pagers. The most abstract aspects of the system, the software agents themselves, were successful in a functional sense. People created plenty of agents and made trades. But the particular technologies of Kasbah did not effectively aid people in building a relationship of understanding with their agents.
