Digital Libraries- Avoiding Unintended Consequences

We digital library developers don't get up in the morning wondering how we can ruin the lives of our patrons. Nonetheless, unintended consequences of our work may damage the capacity of libraries to serve their clienteles, from public library patrons looking for books to researchers searching archives.

We're not the only field that may fall victim to unintended consequences. For an overview of such dangers, see the paper 'The Unanticipated Consequences of Technology' by Tim Healy.

The convenience factor
Anyone who has worked on a reference desk is familiar with this: most of our clientele would rather stand in line to use a computer-based periodical index than to use a more appropriate print index. As more material goes online, this problem gets worse, as patrons may be tempted to do it all from home.

Students already believe that they can get all the information they need to write most reports from the Internet. Sometimes -- and increasingly so, often due to our own efforts -- they are right. Only instructors who require the use of print materials are likely to force students to enter the library.

So how can we combat the tendency to use what is easy over what is often better? One strategy would be for libraries to make print materials much more enticing. Most library catalog records for books offer such miniscule descriptions that it is surprising that anyone enters the stacks at all.

An author, title, and a few broad subject headings are a beginning but only that. What about the table of contents and index for nonfiction books? How about taking a cue from Amazon.com and add jacket covers, reviews, and ratings? The more information we provide about the books we have, the more interesting they will be to those who may have otherwise ignored them. (See Steve Coffman's original idea 'Building Earth's Largest Library' and the response to it.)

The 'act globally' phenomenon
Internet users not only think globally but also act that way. Some seeking library books will prefer a catalog of a large library -- Harvard, Berkeley, the Library of Congress -- rather than their own neighborhood branch. They search the catalog, find one or more books they want, and then request them from that library. In most cases this means they will be referred back to their local branch to request the items via interlibrary loan.

But wouldn't it be great if we could create one place users could go to search the holdings of their local library as well as all others? This megacatalog (did someone mention OCLC?) could automatically highlight local materials via zip code. The basic idea is that we need to get proficient at using technology to allow users to think globally (as in one place to search) but act locally (as in checking out books). Again, see Coffman.

If we want to get really good at this kind of thing, we must get serious about providing opportunities for co-branding (see Digital Libraries, LJ 12/00). Co-branding is much more than simply attaching your name to something. It is also a way to offer local services while a user is viewing remote resources (in the example above, a zip code entry should prompt a local library link).

The keyword search hegemony
Due to the popularity of web search engines, nearly all of which offer one keyword search box to search among millions (and soon billions) of items, users think they can find everything on a topic with a few well-chosen words. Why, they think, should they be forced to understand things like field-restricted searching and controlled vocabularies?

Why, indeed? Our search systems are still too primitive to do the behind-the-scenes processing needed to handle these kinds of searches well. So instead of making some informed guesses about whether they entered a call number or a subject, we make them tell us.

We should be making computers do this. For example, imagine a library catalog in which the person must only enter a few keywords -- a person's name, or a few words from a title, or the subject of a book. Names are fairly easy to detect if you compare the search words against a dictionary, limiting the search to appropriate name fields automatically.

If the search is not a name, perform a combined title and subject search and determine the top five subject headings from the returned records. Present those subjects to the user as links, with the question, 'Is one of these topics what you are looking for?' By clicking on one of the topics, the user would be using controlled vocabulary terms limited to a particular field in the record, all without knowing anything about the underlying complexity behind it.

We must get much better at meeting users where they are or want to be -- and it isn't picking an index from a long list of title, author, or subject. To paraphrase Herb White, 'Only librarians like to search; everyone else likes to find.'

Ignorance is no defense. Errors of omission can have as substantial a set of consequences as errors of commission, so we must try to understand the unintended consequences of both what we do and what we fail to do.

Roy Tennant (roy.tennant@ucop.edu) is Manager, eScholarship Web & Services Design, California Digital Library. He is founder and manager of the electronic discussion lists Web4Lib and Current Cites