|
PROBLEM
The California Digital Library (CDL) website was launched in January 1999. We found that the existing CDL Home page gives an unnecessarily high visibility to the administration of the site, and, as confirmed during our task analysis, misses opportunities to better communicate the scope and variety of resources available for research in this "10th library." While potential users of CDL services range from librarians, professors, and students to university administration and the general public, our focus for this project was on academic researchers, primarily students and faculty, using the CDL as a comprehensive means for accessing online information resources affiliated with the UC System.
Our goal this semester was to redesign the CDL interface to improve the ability of these users to understand the scope of the CDL collection and to facilitate more effective searching across its collections of sources. Specifically, we sought positive responses to the following questions:
Does a user...
SOLUTION OVERVIEW
Our prototype concentrates on orienting the user within the complex CDL universe by bringing more information on the nature and structure of the resources available to the front pages of the website, as well as illustrating that the user should expect to perform a two-level search in order to find specific information within the CDL. The final prototype is a working website of 11 pages--4 main pages and 7 placeholders. The placeholders illustrate the overall proposed information architecture of the site and, once provided with real content, would enhance the full functionality of the site. We have focused our attention on the clarity of the main source selection process rather than on these other aspects of the site, attempting to streamline the user's access to the main research functions and content of the digital library.
In the Browse Sources by Source Type page, we have outlined the types and indicated the breadth of the collections in each source type category (i.e., databases, electronic journals, finding aids to special archives, and reference texts). This page offers users visual aids in the form of icons that illustrate which sources are directly available through the CDL and which are links to independent websites. A separate link at the top of the main Browse page goes to Browse Sources by Topic, and provides a link to the live CDL site's listings of sources by topic. We have not focused on Browse Sources by Topic for this prototype: we found the existing CDL page for this function to be one of its strongest, and believed that clarifying source types was more essential in illustrating to the user what exists in the overall CDL collection.
While shallow compared to the depth of the overall content of the CDL site, our redesign of the front-end website pages offers some viable solutions to some of the frustration and confusion we encountered using the live CDL home and search pages.
Through several iterations of our interface design, we have honed improvements on the original CDL site. The participants in our final pilot study had fewer questions, were significantly less frustrated with the system, and found fewer dead-ends compared with the participants of our original lo-fi design (as well as, anecdotally, compared with the users of the live CDL, who often report it "very difficult" to use). People liked the look of the prototype home page, and found that the source lists gave them a concrete sense of the size and complexity of the CDL resources--a goal we had hoped to achieve.
TASKS & SCENARIOS
At the beginning of the design process we focussed on how a variety of users would use the web-based library system. Our initial tasks therefore sought to encompass the needs of a wide range of users (novice, graduate and faculty) and the typical scenarios these types of users would encounter.
Task Scenarios used for the Lo-fi and Heuristic Evaluations
Scenario #1: Imagine that you are a UC undergraduate student. You are using the California Digital Library for the first time. Rather than looking for a specific resource, you are browsing the system to determine what resources might be available related to your major, Art History. You are interested in both general Art History resources as well as in Renaissance Italian sculpture of the 16th century.We looked to see if new participants would use an overview of site (Site Map) or a guide (User Guide) to begin. And, was our combination of the Search and Browse functions useful to them. Could they learn quickly how to narrow their searches? Scenario walkthroughScenario #2: Imagine that you are a UC graduate student in the Genetics Department. You are looking for a specific journal article that a professor recommended. You don't remember the exact title or reference, but you know that the article was published within the past couple of months, is about mitochondria DNA in sea urchins, and is by somebody named Roberti. Please see if you can find this article.This presupposed some prior knowledge of the system (that could be gained simply from performing the first scenario.) We looked to see if it was obvious how to select a database to search for a journal article, and how to locate the full-text online. Scenario walkthroughScenario #3: You are a faculty member in the History Department studying religious minorities in the 19th-century United States. You are currently working on a research paper about Jews in the American South during the Reconstruction period. It is important for you to stay up-to-date on related developments in your field, so you are using the CDL to look for recent related scholarly journal articles.This scenario tested whether our immediate jumps into the system would work. And, whether using the Profile function was clear or presumed prior knowledge. Scenario walkthrough
Over the course of the testing and design phases, we realized that our scenarios were too broad and required too much prior knowledge. We rewrote the scenarios for the pilot usability testing process to better fit the constraints of the testing time frame and the users' domain of knowledge. However, we ensured that the important tasks were still represented in the new scenarios.
The original scenarios #2 & 3 from our lo-fi testing were designed to test two specific, important tasks: (1) the known-item search; and (2) the comprehensive search over a very narrow topic, e.g., a user needs to know everything published about such-and-such in the past year.
Scenario #1 and #2 are now designed to cover the task of a known item search and also used to test the users' understanding of the system -- not just a two-level search (first identify the database, then search for the specific item), but also a two part search (each with two levels). That is, first use a database to find the citation of the specific item sought; then find the actual item, using either an online electronic journal or the physical library (and its on-line catalog). Of course, we didn't require that the participants go to the physical library, but it's important to consider it a part of the process, and to examine whether users consider the digital library a substitute for the physical library or an assistant in using the physical library.
Scenario #1: Imagine that you are a UC Berkeley student writing a comparative literature paper. You want to compare one of your class readings with Dr.Seuss' famous book Green Eggs and Ham. You want to find a copy of this book at UCB. Scenario walkthrough
Scenario #2: Imagine that you are a UC faculty member studying 20th century theatre history. You are looking for a specific journal article that was recommended by a colleague; unfortunately, you are unable to remember the exact citation. You know that the article concerns the discovery of two "lost" puppet (marionette) plays from the Spanish Civil War (1936-39). The article was published in a leading English-language theatre research journal in late 1997 or early 1998 and was written by J McCarthy.
Please attempt to identify this article. If it is possible, you would like to obtain the full text of the article on-line; if not, you would like to identify libraries where you could view the physical journal. Scenario walkthrough
Scenario #3: You are a UC Berkeley graduate student in the School of Information Management and Systems. You are expected to write a report on the National Information Infrastructure (NII). You want to find out about the history of NII and where it is headed. Your arguments about the future must be supported by past and current trends seen in the information industry. You have already browsed the WWW and now you want to see if there are any published materials in this area.
Please identify a few (top level) CDL sources that might be relevant to your research on this topic. Then choose a couple of these sources and delve deeper to find specific resources and papers on the history and development of the NII. Scenario walkthrough
Sketch Stage
At the Sketches stage of design, we utilized a "Gateway" metaphor (rather than "Library") to help create a mental model of the CDL with less emphasis on it as a storehouse of texts, but rather as a starting point with links to various external information sources. [Sketch of Home] We felt that our Gateway page would help to funnel users into the Collection of resources. From our task analysis, we also decided that a Site map of the CDL website would be imperative to illustrate an overall view of the scope of the CDL collection and site. However, while the need for this type of view was and is still necessary, we found that implementing it in this way was not effective. The metaphor did not help but rather hindered the process of using the digital library to find information in the sources. [sketch of "gateway" search]
Lo-Fi Stage
During the Low-Fi design phase, we pulled a lot more of the Resource types into the Home Page. We included many of the links from the existing CDL homepage, giving equal weight to a list of CDL services and a list of representative information sources. We found that none of our test participants used the services links, but that more weight should be given to the actual resources. Our resource list included a combination of specific resources and broad resource types, and this proved to be confusing to users. Yet it raised our awareness of the users' need to understand the two-level search: first select a source, then search on the source contents. Participants in our testing tried to search content directly and became very disoriented within the site during the search process, because of their mismatched mental model of the CDL structure. We knew we needed to provide them with lots of help, but this proved to be too much information with conflicting and unclear differentiation as to the type of help (quicksearch, help, find...)[lo-fi Search/Browse] [lo-fi SiteMap]
Hi-Fi#1
From our Lo-Fi testing we decided to change a lot for our first Interactive Prototype. The interactive prototype became a working website of 11 pages. The main pages with significant workable content are Home, Select Sources, and Browse by Format. The Select Sources and Browse pages function directly with the live CDL website to retrieve sources or source lists (i.e., search results.) The use of color was used for showing some general differentiation between sections. Icons were created in the Browse page to indicate whether there were more sources and where they would be located (on CDL or linked elsewhere) to help the user maintain status of their whereabouts in relation to the site.
The information that was presented on the Hi-Fi#1 Home Page more coherently presented the scope of the TYPE of resources to the user. Since we found that users went in circles during search sessions and virtually ignored "Services" and detailed explanations, we tried to graphically force a sense of the scope of Sources on them by presenting the types of Sources upfront. Throughout the main pages we revised the terminology with more natural language and less librarian jargon, e.g., "Format" was changed to "Source Type", "Archival Finding Aids" changed to "Finding Aids to Special Archives", and "Electronic Journals" changed to "Electronic Journals on the Web" (to differentiate this from the many abstract journal databases that the UC system accesses.) We also eliminated detail of "Services" from the Home page.
Hi-Fi#2
The Heuristic Evaluation performed on our Prototype #1 resulted in our implementation of the use of Frames in HiFi #2 to keep the navigation bar always visible especially during the long Browse Sources by Topic lists. The navigational bar itself became a mini-site-map to clarify one's position within the website. And the Home Page was also reformatted to give more emphasis on the Sources and entry points into the site and less screen real estate to the marketing blurb for one thing. We added descriptions to the site links to the front page and once again finessed the terminology to help the user understand what and where they would be going to on a click, for example, "Select Sources" became "Choose a Source".
Hi-Fi#3
The pilot study on our HiFi #2 identified a continuation of one of the biggest areas of trouble: how to give the user a good understanding of identifying valuable sources from long browse list, yet avoid overwhelming them. The terminology used does not always afford the user a correct "guess" as to the content of that source. Users tended to select an inappropriate database for example because of the similarity in titles between Journal Article Indexes Database and Electronic Journals on the Web. We had explained the differentiation on the Home page, but realized that our layout of left/right title to explaination did not adequately afford scanning of key concepts. We have since reformatted this table to better facilitate a quick scan--top to bottom--of the types of sources by creating a tighter graphical integration between links and descriptions. We also reformatted the Home page for HiFi#3 to offer shortcuts to popular starting points.
Most Valuable Evaluation Technique
Low-fi prototyping limited the accurate portrayal of our design choices (e.g., our search button on the home page was not always recognized as a functional button); nonetheless, we seemed to have learned the most from the lo-fi prototype testing, as evidenced by the number of changes in our design of the first Hi-Fi prototype. The lo-fi prototype was tested in combination with the CDL live site, and this afforded us the opportunity to test the effectiveness of whole site and the users' expectations and understanding of the digital library. We found that what we had implemented in our lo-fi was not used enough, and that users moved too quickly into a struggle with the existing search and results of the live CDL site. We therefore concluded that the main pages in our redesign had to somehow offer more help and contextual information to support the user's search process. This significant change to our design vision was implemented in the Hi-Fi#1 stage and fleshed out as the design iterations continued.
FINAL INTERFACE
Description of the Final UI Design
We have created a working website of 11 pages -- 4 main pages and 7 placeholders. The 4 main pages:
are functioning pages which we have filled with content and tested throughout this design process. It is important to note that at this final prototype stage, the Browse Sources and Choose a Source pages function directly with the live CDL website to retrieve the actual sources, source lists, and "more info" on sources. Consequently, the user will inevitably encounter live CDL pages containing UI problems that are outside the scope of our interface redesign project.
The Home Page shows the user that there are several source categories (databases, electronic journals, finding aids, and reference texts) and some services that the CDL provides. The most significant change from the testing of Prototype 2 is that we have removed the reduntant descriptions of in-site links and instead provided starting points for searches of the more popular databases directly from the home page.
Because of the links to the real CDL site, a user of our prototype can successfully search for a topic-specific source. The real CDL site allows users to search for a source by topic ("Browse") or by keywords ("Search"), but not both. Our design unifies these two options; however, because we link to the existing CDL's cgi script, we are limited to implementing only keyword searching without functioning topical searches (The list of topics includes an explanatory note). We would also have prefered to implement multiple choice selections for both Source Type and Topic, but the live CDL site only allows for radio-button single selections at this time. Had we had the opportunity to build our own databases and the skill to implement our own search engine, we might have been able to implement this UI design without the crutch of the live CDL site. Overall, we felt there was more of a need for the user to understand the breadth of the Sources supported by CDL, and therefore, have chosen to focus on building the Browsing and general source selection page rather than on this "search" page.
Remaining Prototype Problems/What was Left Unimplemented
Our pilot study was also partially affected by a few technological difficulties. We chose to implement keyword searching rather than topical searching. One participant used the Keywords search on that page, although she was fairly sure that her query would not be relevent (the result supported her assumption). In the other cases, participants used the Topics search (despite the note saying it is not functional in this prototype), retrieving a CDL error message and a new search form.
We also have few unsolved problems with frames. Since each in-site link needs to open a new document in both frames, one muct hit the back button twice to go back a single step. We suspect that there is a simple way to solve this with JavaScript, but it remains to be implemented.
Also, if we had the time (and the skill), we would generate our browse tree automatically rather than using multiple individual HTML pages. This could improve the speed and smoothness of the browse function.
"Wizard of Oz" Techniques That are Required To Make it Work
Essentially the interface works without divine intervention. There are four caveats to this:
Tools Used
Dreamweaver was the main tool to create functional webpages. We also used PhotoShop for icons and a GIF creator for part of the navigational sidebar. Each page was created by hand, which proved to be tedious. Had we had more time to become more productive in using an application like Cold Fusion to link a database to the HTML, we might have been able to generate the multiple Browse pages with a database backend, rather than building each by hand. Overall, though, using these simple applications proved to be easier for implementing our vision of the website than were the paper sketches from the lo-fi prototype.