A general goal of my research is to understand how people use technology to seek and use information (e.g., Marchionini, 1995). More specifically, I have focused on end-user information seeking rather than professional intermediaries and on primary (e.g., full-text) rather than bibliographic sources. Studies of children using electronic encyclopedias, domain specialists (e.g., computer scientists, economists, attorneys) using full-text and hypertext systems, and instructors and students using a hypermedia corpus demonstrate that end-users prefer browsing strategies to analytical strategies, are concerned with answers rather than queries, and of course, vary enormously in their abilities and attitudes. My colleagues and I at the Human-Computer Interaction Laboratory have applied these theoretical results to the improvement or original design of library interfaces for end-users such as the National Library of MedicineUs Grateful Med and the Library of CongressU ACCESS system (Marchionini, Ashley, & Kortzendorfer, 1993). Likewise, these results have influenced the evolution of the Perseus hypermedia corpus, a specialized digital library on the ancient Greek world (Marchionini & Crane, 1994). We have recently entered into a cooperative agreement to work with the Library of Congress (LC) to develop the interface for the National Digital Library.

The approach to design we will take in this project builds on our experience in previous work. It may be characterized as a workgroup approach that involves users at multiple iterations of the design process. It is similar to what HCI designers call participatory design and what architectural firms call workshop design, each of which actively engages users in developing and testing products. The workgroup approach is somewhat more constrained in two ways: first, it includes a preparatory phase in which the work group is oriented to the perspectives and research experience of the HCI experts; second, the nature of libraries leads to a cascading set of user communities that have different roles in the design process.

The preparatory phase includes a set of sessions in which different members of the design team present a seminar on an aspect of HCI theory or practice. In the LC project, topics such as: information-seeking strategies in electronic environments, usability testing, interface design methods and procedures, and visual interface trends for information systems, are planned. These sessions serve to establish shared vocabulary and illustrate some directions for the ongoing work.

There are at least four sets of users in libraries that are oriented to public service: library staff, frequent and sophisticated users, occasional users, and potential users. Library staff are most dramatically affected by systems since they both use these systems themselves and teach or help other users to use these systems. Additionally, these users are "on-the-job" constituents and thus are able to participate in sustained working group efforts. Frequent or sophisticated users in libraries are those patrons who regularly use library services and have a vested interest in learning to use systems that meet their information-seeking needs. Since these users are not employed by the library, their participation is voluntary and possibly intermittent (in the case of the National Library of Medicine design we used physicians and researchers in this group as usability testers for an advanced-stage prototype). Occasional users are traditionally invited to participate through verbal interviews or written questionnaires in the library but in a digital library setting, online questionnaires and electronic feedback forms may be used to assess their needs and to gather their reactions to the system. Most challenging are the potential users, since they do not make themselves known to the library. In traditional settings, community surveys may be used to determine why these people are not users and what services might encourage them to become users. In the digital library, these users include "surfers" who stumble onto the site as well as all the people who do not use the library or computers. We are exploring ways to solicit perceived needs of non-users and determine what conditions would lead them to become users. One approach is to conduct verbal interviews in public places; another is to use written questionnaires in non-library settings (e.g., a day-care center) another is to add brief online surveys to web sites and other existing services beyond the library.

Ideally, the working group would involve participants from each of these communities, but in actual practice, working groups are composed mainly of the library staff with intermittent participation by regular users. The orientation sessions lead to the group gathering and discussing existing interface examples and first-phase specification of functions and information resources to be included. These specifications are then used to sketch screen displays on paper (quick drawings, sticky notes, plastic film, different colored markers are used to differentiate options or represent active mechanisms) that are the basis for critical discussions and revision of specifications. Meetings are devoted to brainstorming and revising the mock-ups until a design emerges that can be mocked up with some tool such as Visual Basic. For the LC project we are also using a tool (IVEE) developed to rapidly prototype dynamic queries (Ahlberg & Shneiderman, 1994). The mockups are then critiqued by the group and as many iterations of revision are completed until a consensus forms. This prototype then is used as the basis for user testing with those beyond the working group.

During this process, the HCI team uses research results and design experience to guide the design toward a realistic and sensible conclusion. Since the logistical environment forces us to work mainly with one user community in the working group, it is important that salient research results from studies of end-user information seeking be considered, lest the experience of experts lead to biased designs. Of particular importance are results such as:

• End users are answer-oriented rather than query-oriented;
• Boolean-based systems are difficult for casual end users and they prefer "natural language" interfaces;
• End users use very few terms in their queries;
• People hate to wait (500-1000 ms response limits; and
• People are easily overloaded with information.

Throughout the entire process, design specifications are developed and formalized so that implementation is driven by the results of the user-centered design rather than by system constraints. In addition to these results, designs for library systems are influenced by the nature of library collections. Important characteristics include:

• Persistence in the collection;
• Collection is driven by policies (additions and deletions);
• Need to support within object as well as across object searching;
• Generally large volumes of information;
• Zipf-like access patterns (intermittent, specialized);
• Standardized pointers and representations for collections and items;
• Disparate user populations that in public-oriented libraries include the full range of Human capabilities and characteristics (knowledge; skills; abilities; attitudes; styles, preferences and inclinations; experience (expertise in domain); ages; and special needs.

In essence, the HCI team must help the working group understand not only what system features are possible, but that there is a design space defined by users, tasks, and systems.

References

Ahlberg, C. & Shneiderman, B. (1994). Visual information seeking: Tight coupling of dynamic query filters with starfield displays. Proc. of ACM CHI94 Conference, (April, 1994). 313-317 + color plates.

Marchionini, G. (1995). Information seeking in electronic envrionments. NY: Cambridge Univ. Press.

Marchionini, G., Ashley, M., & Kortzendorfer, L. (1993). ACCESS at the Library of Congress. In B. Shneiderman (Ed.), The sparks of innovation in human---computer interaction, pp. 251-258. Norwood, NJ: Ablex.

Marchionini, G., & Crane, G. (1994). Evaluating hypermedia and learning: Methods and results from the Perseus Project. ACM Transactions on Information Systems, 12(1), 5-34.