Let patrons evaluate materials - in you automated catalogs. A daring proposal from a leading educator.
THE CONTINUING EVOlution of information technology will soon enable, and indeed require, changes both in the design of information systems and in the way librarians define their roles. Why will these change take place? The answer lies in the convergence of technological know-how and patron need. Technology has now developed to the point that we can dramatically improve our library information systems' selectivity and performance. Increasingly, patrons will expect from librarians and information systems advice about the authoritativeness, or intellectual worth, of material by making available user-supplied data (USD) - evaluative comments from other users about specific holdings.
Software capable of such feats is rapidly evolving - the buzz word is hypertext. What I propose is simply that 1) access mechanisms to bibliographic information and to full text will gain hypertext-like flexibility, and that 2) USD will dominate the data generated by libraries or publishers in those systems.
Cynics might say that we must needs develop USD-friendly systems because library users will want them whether they need them or not. In fact, our patrons need such systems desperately.
The information explosion has continued unabated, despite Derek Price's 1963 prediction (1) that the exponential growth of scientific literature would have to level off, probably in the 1970s. Not only has that never happened, but there is no indication of it leveling off soon. (2) Conversely, Brian Aveney has speculated that the growth of information technology will cause information to mushroom even more rapidly. (3) Perhaps Aveney's theory explains why Price's prediction failed. In any case, today's debate is not over whether the information explosion will continue, but merely about its growth rate.
But while the information explosion continues to grow, our ability to selectivity retrieve is not keeping pace. True, a parallel information controllability explosion has made us better able to store, transmit, and selectively retrieve appropriate material in a Boolean sense. (4) But paradoxically, we have created a new problem for ourselves by achieving some measure of control. We can retrieve more articles than ever based on a concept, a keyword, or combinations thereof. But how do we have weigh the likely utility of those articles? In the past, we have relied on library users to browse and scan the retrieved set. But our ability to amass more material during a search has also increased the bottleneck of potentially relevant documents to examine.
Librarians have spent the last third of a century attempting to devise systems that enable their patrons to more effectively judge the relevance or likely utility of materials - with embarrassingly little success, as Cyril Cleverdon's Cranfield test so painfully revealed. (5) We can certainly hope to make more progress in that direction; there is real potential for the library application of artificial intelligence techniques, but progress will not be easy and it will not be quick. (6) What we can do with relative ease and effectiveness is to satisfy our patron's need to assess the authoritativeness of documents by adding hypertext capabilities to library information systems. Then, users could input evaluative data directly.
Library patrons want authoritative information. This surely is not news to most librarians, nor to students of scholarly or other communications. A 12-year-old may want the straight skinny from Jane's All the World's Aircraft, while a high school student may seek six "good" articles on Ukrainian nationalism. A graduate student may want to know whether William Wimsatt is still considered the authority on Alexander Pope, and a researcher may need to determine the most authoritative hypotheses about the probable relationships between cytokines and thyroid epithelial cell growth.
Researchers frequently glance at an article's list of references to assess its authoritativeness before deciding whether to invest time in scanning it. An eminent physician described a provocative example of this practice in the early 1980s at an invitational seminar on information systems at New York Public Library sponsored by the Rockefeller Foundation. Recounting how he typically used MEDLINE, the doctor told of one occasion when he searched the online database to determine whether a patient's jaundicelike symptoms might be drug-induced. The salient and rather sobering point is that he decided to alter the patient's drug regimen solely because he noted that the title or one retrieved article implied that a particular drug might trigger hepatitislike side effects. The doctor consulted neither text nor even abstract before making that important clinical decision. Instead he relied quite consciously and quite deliberately on the fact that the article had appeared in British Medical Journal as a sufficient indicator of authoritativeness (IA).
All authoritativeness data are in essence supplied by users of that data, whether the IA be citation counts or Nobel prizes. USD is the deal mechanism with which to supply library patrons with the IA and relevance date we know they seek. We can and should design system to encourage users to input their own value and authoritativeness judgments. In the electronic era, there is no reason not to encourage such user input, given a system adequately designed to support USD.
When we were limited to print on paper, librarians certainly could not encourage patrons to scribble marginalia in library books or on catalog cards. That was defacing library property, in many cases a prosecutable misdemeanor. And we certainly did not encourage users to add their own cards to our catalogs.
There were good reasons then for discouraging USD in library collections. For one thing, it would have been difficult to distinguish between defacement and editorial commentary. For another, there was no practical way to delete USD from a holding or catalog card for patrons who preferred their source materials in pristine condition. Of course, a library could have maintained two sets of holdings - one clean and one marked up, but the benefits would have hardly outweighed the expense.
In an automated environment, these difficulties disappear. A patron can choose what version to access: clean, replete with USD, or with selected USD.
Information systems can make USD available in a number of ways. The first and most obvious is by supporting the electronic equivalent of marginalia. In effect, we should link our bibliographic and text retrieval systems to our electronic bulletin boards - in other words, mesh library catalogs and full text databases into hypertext systems in which USD comprises a major component of entries. Ideally, such a hypertext environment would allow one to follow the USD of a user whose insights one appreciated, just as one might follow the review of a favorite and sympatico movie critic.
We could extend the concept by programming systems to solicit USD. Hypertext systems might also display aggregate circulation date - while safeguarding user privacy, of course. Patrons might benefit by knowing what materials other users choose most often, or how often an item has been put on reserve and for what academic courses. Other applications include linking reviews with materials; offering citation, co-citation, and co-reference links; (7) and making available data on which articles users have retrieved most often in response to searches.
In fact, there is a vast amount of user intelligence that we unnecessarily discard besides IA. Another long-unutilized resource is the "or-group" thesaurus users create during online searches.(8) Why not maintain online a ranked list of what terms searchers have linked in Boolean "or" statements? Thus, once I had structured a search about South Africa by stringing together the terms apartheid or discrimination or disinvestment, subsequent searchers researching apartheid might discover disinvestment on my "or" list and perhaps add a term to their search they might otherwise have overlooked.
In short, we need to change tour mindset from that of creators of catalogs for others to use, and for which we are exclusive suppliers of all the data, to providers of a catalog users are free to supplement.
There are three reasons why we haven't implemented such a system to date. First, although technically speaking implementation would be relatively easy as a pilot project, it would be neither easy nor inexpensive to implement at the operational level. ;Needed would be extensive software development, a good deal of coordination, and substantial resources. To facilitate more manipulation of more data requires more gigabytes of memory and more mips of computer power. We are only now at the stage we can realistically anticipate such systems becoming cost-effective.
A second roadblock is simple professional inertia. For what once were logical reasons, we have simply never thought about recording USD, We have captured the insights of authors and publishers, indexers and catalogers-but never users.
Third, and by far the most disturbing, is our cultural blind spot toward USD. Though information overload and system enhancements have been staple subjects in library literature for years,(9) we have almost never mentioned getting a better handle on IA. For instance, Tefko Saracevic presents in a classic study a clever Chinese menu style list of definitions for relevance-the concept of distinguishing what in a retrieved set is useful to the patron. The definitions include precision, pertinence, quality, timeliness, and aboutness selectivity. Striking in its absence, however, is the word authoritativeness or any synonym. Pertinent and relevant as authoritativeness is to library users, we librarians have simply ignored it.
Why? We see IA as antithetical to the role we have cast for ourselves as defenders of First Amendment rights.
Professional blind spot
Librarians have always taken pride in opposing censorship and in supporting an open marketplace for ideas. A perennial struggle in librarianship is between the exercise of selection and the eschewing of censorship. The "compromise" we believe we have achieved is to represent that open marketplace by selecting the best representations of those myriad ideas. Thus, we expect ourselves to be nonjudgmental about the validity of ideas in the marketplace, and yet exercise judgment in selecting materials that best represent those ideas. Often, that ambiguity is lost or at least obscured precisely because we tend to perceive librarianship as inherently objective.
So, librarians may instinctively see the design of information systems to support access to IA as a violation of all that we hold near and dear. It is not. Logically, what is the difference between an opinion expressed in a book or in our catalogs? First Amendment rights should apply equally in either place.
In making their historic compromise, librarians have cast the catalog in the role of neutral descriptive tool (though Sanford Berman reminds us of how many cultural judgments slip in unobserved(11). That neutral role is just an artifact of the printed card catalog. There is no reason why the catalog cannot contain judgmental data too, so long as it is clearly labeled as such. It is the access that must be neutral, not the data.
If we are going to serve our users well, we must overcome that blind spot. One way is by thinking in terms of USD-a concept which shouldn't conjure up any library demons-rather than authoritativeness data. As file users, the typical patron may be seeking authoritativeness in manipulating USD. But as system designers creating a file, librarians would simply be providing, not endorsing, the opinions expressed in USD.
To design systems effectively, we must understand how patron interest in IA. Remaining oblivious will surely cripple our system design efforts.
Access to authority
Librarians will have to rethink how they, as system operators and designers, interact with the information their systems purvey. One potential conflict is that between providing USD and safeguarding the users; right to privacy. For example, students may find it useful to access their professor's USD and use pattern. That professor may well authorize the former, but probably not the latter. Should that professor even have the right to authorize casual access to personal use pattern? Could someone's disinclination to authorize such access ever play a role in a latter-day McCarthy witch hunt?
There are also substantial ramifications for interlibrary cooperation. The value of user intelligence and USD is not geographically limited. A student at Rosary College may well want to know whether a title is frequently used and often put on reserve at Yale University. USD applications will reach their greatest potential in an atmosphere of organized planning and interinstitutional cooperation.
To take full advantage of the opportunities USD offers, librarians will have to do a great deal of coordination and planning. For example, what data can we share while continuing to safeguard privacy standards? We are all familiar with the invisible college phenomenon, in which a researcher in, say, information retrieval would find far more valuable the data generated by others researching the same topic at other institutions than data supplied by colleagues on campus whose focus is different. Ultimately, the provision of USD-though eminently useful and realizable on a local level-is another inroad toward the deinstitutionalized library.
The pendulum has swung from the centralized cooperative library systems of the 1970s back to local systems. In fact, there is now substantial concern that the advent of cheap cataloging data on CD-ROM may cripple the goose that laid the golden eggs of library cooperation-the bibliographic utilities. (12) However, the cross-institutional utility of USD may swing the pendulum back again.
We are on the brink of realizing a technology capable of supporting USD. To make the most of it, we must expand our exclusive pattern of offering only data we select to an inclusive practice of incorporating USD too. Even more of a challenge, we must tear down the cultural roadblock that prevents our offering judgmental and evaluative information in our bibliographic access mechanisms. Finally, we must recognize that this opportunity necessitates an entire new round of library information system planning.
( 1.) Price, Derek. Little Science, Big Science (New York: Columbia Press, 1963). ( 2.) Griffith, Belver. Personal communication, November 1988. ( 3.) Aveney, Brian. "Electronic Publishing and the Information Transfer Process," Special Libraries (October 1983:p/ 338-344). ( 4.) Koenig, Michael E. D. "The Information Controllability Explosion," Library Journal (Nov. 1, 1982: p 2052-2054). ( 5.) Cleverdon, Cyril. "Cranfield Tests on Index Language Devices," ASLIB Proceedings (June 1967:p. 173-194). ( 6.) Salton, Gerald and Chris Buckley. "Parallel Text Search Methods," communications of the ACM (February 1988: p. 202-215). ( 7.) Yermis, Ira. A Citation-based Interactive Association Retrieval System. PhD dissertation, University of Pennsylvania, Department of Computer and Information Sciences, 1975. ( 8.) Reisner, Phyllis. Evaluation of a "Growing" Thesaurus (Yorktown Heights, N. Y.: IBM Watson Research Center, 1966). IBM Research Paper RC-1662. ( 9.) Warren, Kenneth S., editor. Selectivity in Information Systems: Survival of the Fittest (New York: Praeger, 1985. (10.) Saracevic, Tefko. "The Concept of 'Relevance' in Information Science: A Historical Review." In Introduction to Information Science. Tefko Saracevic, ed. (New York: R. R. Bowker, 1970). (11.) Berman, Sanford. Prejudices and Antipathies: A Tract on the Subject Heads Concerning People (Metuchen, N. J. Scarecrow Press, 1971). (12.) Molholt, Pat. Library Networking: The Interface of Ideas and Actions (Washington, D.C.: Department of Education, Office of the Assistant Secretary for Educational Research and Improvement, Library Programs, June 1988).…