IBM EXPO '92 Guest Services System
J.F. (Jeff) Kelley,
W. Bennett, S. Boies, C. Cesar, J. Gould, S. Greene, L. Jones, J. Kesselman, R. Mushlin, S. Spraragen, J. Ukelson, C. Wiecha
IBM T.J. Watson Research
30 Saw Mill River Road
Hawthorne, NY 10532
Contact J.F. Kelley
This paper was presented at the Annual Meeting of the Human Factors and Ergonomics Society Meeting in 1993.
There are videos available of the project at the IBM Expo 92 Webpage.
In 1994, the author demonstrated the system to Senator John Glenn at the World Summit on Trade Efficiency in Chicago, IL.
The ITS group at IBM's T.J. Watson Research Center is a multidisciplinary group that has developed a new approach to application development. As part of making our tools better, we use them to develop challenging applications. One recently publicized example was an integrated set of multimedia applications developed for the EXPO '92 World's Fair in Seville, Spain and used by more than 10 million visitors. . This application allowed the visitors to "finger-walk" the site, make restaurant reservations, exchange multimedia messages, read the daily news, make fingerpaintings, have their opinions surveyed, and more. The system, distributed throughout the site, was a high-speed network of 231 multimedia public access kiosks. The system comprised high-resolution touchscreen displays, thousands of images showing aerial photos and pictures of pavilions, over 100 multimedia country stories, video cameras, audio input and output, and badge and ticket readers. This paper is an overview of the Expo kiosk application as well as the ITS tools and user-centered, iterative development process that made it possible.
In 1996, the ITS group won the Alexander C. Williams Award from the Human Factors and Ergonomics Society. This is the highest award given in that profession and is directed to "the outstanding contribution of human factors to the design of a real-world application".
This application is best described by the video: IBM EXPO '92 Guest Services System (available from author). Portions of this video were shown at a meeting of the Metropolitan Chapter of the Human Factors Society in March, 1993 and at the Human Factors Society Annual Meeting in Seattle in October, 1993. The video has 4 parts:
1. Overview (17 min)
2. Hardware (2.5 min)
3. Central Kiosk Control (3.5 min)
4. User Interface Design (9.5 min) The sequences in the video showing use of the system are all of unrehearsed, real visitors at EXPO '92. One of the kiosk workstations was instrumented with extra video cameras which, combined with a portable video camera, captured actual use of the system. While pains were taken to be as unobtrusive as possible, no attempt was made to conceal the recording; however, users seemed to manifest very little awareness or concern about the presence of the recording devices.
The 1992 World Exposition, EXPO '92, which recently closed in Seville, Spain was a six-month event designed to showcase over 100 country pavilions, as well as a number of technology milestones. IBM, as the official information service provider for EXPO '92, contracted with EXPO '92, S.A. to provide a public use information system. The Guest Services System was defined, developed, installed, and operated by the ITS group at IBM Research working in cooperation with IBM Spain and EXPO '92, S.A.
The Guest Services System is a very large distributed multimedia platform that provided EXPO '92 guests with a wide range of services. These included information about EXPO '92, the countries at EXPO, and the contents of the pavilions, an on-line newspaper that was updated daily, a multimedia electronic messaging system for use by the guests, an on-line restaurant reservation system, and a finger-painting facility. The Guest Services System was, by any measure, a resounding success. Between 10 and 15 million visitors were estimated to have used the system, which registered over one billion interactions. The system was extremely reliable, registering over 99% availability during the six months that the fair was open for visitors (from 9 am to 4 am, 7 days-a-week).
The Guest Services System represents a number of technological innovations both in user-computer interface design and systems design.
Typically a kiosk has been thought of as a stand-alone device. Past information systems at world's fairs have not been connected in any way. This has resulted in both function and maintenance limitations. The Guest Services System was a collection of over 300 high-performance workstations cooperating over a high-speed network without the need of a connection to a host mainframe computer. It was a leading-edge demonstration of the function and maintenance advantages that are provided by a network of workstations.
There were 231 public-access multimedia guest stations housed in 33 kiosk structures. Each workstation comprised a 33 MHertz PS/2 model 95 with 16 MBytes of memory and 1.2 GBytes of disk storage. The displays were 19" IBM model 6091's, with touchscreen overlays. The resolution of the displays was 1280 x 1024 with 256 colors. Attached to each guest station was a video camera with computer-controlled tilt and lighting, a microphone, stereo amplifier, speakers, daily ticket reader and a smart-card reader for season passes.
Network, Command Control, Restaurant Workstations, Infrastructure
Each kiosk housed seven guest stations and one server on a double-16 MBit token ring LAN (local area network). The kiosks were connected to 35 additional servers in the CKC (Central Kiosk Control) via a high-speed fiber optic network (at about 30 Kilometers, the largest token ring network in Europe at the time). The CKC was, in turn, connected over fiber to the 25 restaurants equipped with reservation workstations.
The operation of the entire system was controlled by a series of operator stations. These stations were used to monitor the operation of the entire system. In addition, these stations were used to distribute both information and programs to the guest stations and servers.
A number of stations were used for the preparation of content for the system. An audio workbench was used for the capture, processing and preparation of the music segments and sound effects. Several scanning workbenches were used to scan the information for the stories as well as the images associated with the daily newspaper. Numerous story preparation stations were used to organize, review, and revise the stories. A newsmaker workbench was used to create the daily newspaper. A translator's workbench was used to translate the news from Spanish into both French and English.
HCI and Systems Design Innovations
Many interaction techniques were developed to address the ambitious functional requirements and to solve problems observed during the design process. The primary interaction technique used in the system was the "Slide-to-see / Lift-to-do" touch technique which allowed people to preview a descriptions of what they would see if a selection were made. It also provided a clear identification of the touchable regions of the screen while the screen was being touched without detracting from the beauty of the screen images at other times. Video capture and finger-painting gave new users of the system an opportunity to discover the power of this interaction technique, but users who ignored it and simply poked at the screen were accommodated as well since the system operated intrinsically in both modes.
The touch-based interaction technique was also used to drag objects on the screen. An example of this was the finger clock which proved to be an intuitive way for people to specify times by moving the hands of a traditional analog clock with their fingers. This is an example of a very concrete metaphor that gave users a sense of direct control of what occurred on the screen. Similarly, in finger-painting, the user's finger became a "brush" for drawing on the screen. The immediacy of the results was so striking to youngsters that many looked at their fingertips after drawing a stroke as if they expected to see an ink stain.
More important than the individual interaction techniques developed for the Guest Services System application was the way in which the application was designed and developed. Previous experience with the 1984 Olympic Message system and other applications had taught the ITS design team that the key to delivering interesting function to a wide audience is in the design of an appropriate, user-friendly interface and that the key to accomplishing that is to involve representatives of the ultimate user population early in the design process. This goes beyond mere prototyping since prototypes are traditionally discarded once the "real" application development begins and, once the prototype is discarded, involvement of the intended users also stops.
The alternative approach used here was to develop the actual application using a tool base that allows for the iterative evolution of the actual application from very early prototypes, through installation of the functioning system and beyond into maintenance and extensions of function after installation. In this case, input from actual users in a natural environment was collected from very early on. Beginning when EXPO '92 was just a big construction pit, busloads of tourists would come through the site and would encounter the early prototypes of the Guest Services System. System designers would often be standing by, clipboards in hand, ready to take note of the strong and weak points in each design iteration. This evolutionary approach continued up to and beyond opening day. Many significant enhancements to the function and some major changes in the way it was delivered were made during the course of the fair without any interruption in service to the 250,000-plus daily visitors.
Guest Services System Function and Usage
One obvious indication of the success or failure of the Guest Services System and the underlying tools and methodologies it exemplified would be some measure of how much the system was actually used. If the function wasn't both useful and usable, there would be a lot of kiosks sitting idle; this rarely occurred.
During the course of EXPO '92, there were more than a billion interactive events logged by the Guest Services System, making it the most-used multimedia application in the world. In a typical week during EXPO '92, guests used an entry ticket or a season pass to identify themselves to the system over 230,000 times. By the end of EXPO '92 guests identified themselves to the system well over 5.5 million times. In addition, uncounted thousands of guests used the story and news features of the system without using the identification feature. During a typical week guests touched the screen well over 7,000,000 times.
The system had about 100 multimedia stories. These stories used a combination of 5,000 high-resolution digital images, text in three languages, music, and sound effects to tell the guest about EXPO '92, the countries participating in EXPO '92, and the pavilions. These stories represented about 1.5 GBytes of images, music, and sound effects. This very popular and informative aspect of the system was accessed over 175,000 times in a typical week.
The electronic mail feature was a novel and popular aspect of the system. Guests could create a multimedia message using a combination of digital images, music, and voice messages. The guests could use the video capture facility to create a digital image of themselves for use as an personalized electronic signature. Addressing and access was automatically controlled by information encoded on the entry ticket. In a typical week, guests took over 500,000 video images, sent over 32,000 multimedia messages, and received over 65,000 messages.
Guests could express their creativity using a finger-painting program. Depending upon artistic abilities, guests created a wide range of paintings. During a typical week guests selected finger-painting about 231,000 times. Because finger-painting was a popular diversion, it was an effective way for users to get used to touching the screen and interacting with the system.
The electronic EXPO '92 Daily was a another popular and informative aspect of the system. It contained stories about EXPO '92, weather forecasts, satellite weather maps, and national and international news. A key feature of the system was the collection of electronic copies of the front pages of six international newspapers. In addition, a real time feed from the news services provided an up-to-date summary of current events. EXPO '92 Daily had an average weekly circulation of about 48,000. This made it the second most widely read daily news publication in the Andalucia region.
A restaurant information and reservation system provided a means for the guest to select a restaurant and make a confirmed reservation. Using the system the guest could browse the menu of over 20 of the formal restaurants that were on the site. Guests could select a restaurant based on the menu, price, and availability of seating. In a typical week guests accessed the restaurant information and reservation system over 42,000 times.
The usage statistics were very impressive. Without a doubt the Guest Services System has had more users than any other multimedia application. These statistics, of course, are one clear indication of usefulness and usability of the system. However, as impressive as they may be, they do not capture the fun and joy people had with the system. They simply had a remarkably good time using an IBM computer system. The joy was remarkable when guests, of all ages, found that they had caused their own pictures to appear on the screen. There was something really special about seeing a family clustered around an IBM kiosk deciding how they would spend their day at EXPO '92.
The Guest Services System had one primary goal: To enhance the quality of the guest's visit to EXPO '92. We believed that by accomplishing this goal we would also create a showcase for technology, particularly in the areas of user interfaces and multimedia applications.
One design point called for the system to be used by about 30,000 new guests a day. Most of these guests would have little or no computer experience and would vary tremendously in age, cultural and language background, education and expectations. Clearly no training was possible. The daily number of new users and the time constraints made even an on-line tutorial impractical.
A second design point called for the Guest Services System to provide a wide range of function. The system provided a multimedia electronic messaging facility, an on-line newspaper, structured and unstructured searches thru data bases, an interactive video capture facility, a finger-painting application, etc. Implementing a walk up and use interface for any one of these functions would be a challenge. It was suggested more than once that attempting to implement a walk up and use interface for all these functions was unrealistic.
A third design point called for the Guest Services System to look really good. It was clear that guests would form an impression of the system by considering what it did, how easy it was to use, and how well it looked. Although a skilled graphic designer could have created any number of beautiful screens, it would not be appropriate to have the interface merely look good, and unacceptable for it to look good at the cost of usability. Instead, we took the challenge of applying good graphic design principles to not only create visually appealing and inviting screens, but also to enhance the usability through the screen layout, colors, text font choice and placement, and icons.
A fourth design point called for the Guest Services System to be interesting and fun to use. It was simply not enough to make it possible for the guest to send, for example, a multimedia message. It was also important that the guest had an enjoyable time doing it. The development team believes that all interactive systems should be interesting and fun to use. But in a public use information system at a Universal Exposition this was key.
The development team expended a great deal of effort to meet the design points and thus reach the primary goal of the application. The challenge was to direct this effort so that the result was not just the installation of one world-class public-access transaction system, but the development and validation, through actual use, of a set of tools and design methodologies that could be extended to the development of other public-access and desktop transactional systems.
Future consumers of all transaction-oriented, public access, and large-scale multimedia systems will benefit from the results of our innovations. Previous public access applications have tended to provide only simple information browsing functions in a stand-alone environment, with the associated difficulties in maintenance of timely information (and the application code itself). We proved public access systems needn't be limited in this way to disconnected "page-turning" functions, but that interesting, engaging, transactional functions can be successfully implemented in walk-up-and-use systems. Networked transactional systems also allow for delivery of timely information (and a facility for application code maintenance). Untrained users can be empowered to create as well as consume information as long as the interface is correctly designed for ease-of-use. We demonstrated that the correct way to design user-friendly systems is to follow the principles of user-centered, behavioral, iterative design. We were successful because we involved end-users as partners at the very beginning of the design process and, with reference to feedback from continuing use of prototypes in a natural setting, we evolved the application incrementally. We hope that this message will not be lost on the developers of other large-scale public-access applications.