Professor
Jeffrey K. MacKie-Mason
Department of
Economics and School of Information
University of Michigan
Of course, library and information science research is not a new field. Some of the inputs to the production of library services (processing and communications) are now cheaper. So what? Why does this justify a new research programme? Why not continue as before, merely acknowledging with pleasure that the job is getting easier? Not surprisingly, I think there is something new going on, that justifies a new research programme. However, I think the cautionary note is warranted. If we want to do good science, science that is more worth funding than the many interesting projects on other topics that cannot be funded each year, then we need to repeatedly ask (and try to answer!) the questions: What unsolved problems are there? Why are they important?
Let me offer a partial, but incomplete answer. This observation reflects my primary training as an economist, which means it is personally important but not necessarily the most important new challenge. The observation is this: A primary consequence of the dramatic fall in sand and silicon costs is the change in relative scarcity of inputs to libraries and federated information systems. As a result, everything is different.
Design, implementation and management of information systems all require choice among alternative configurations of costly inputs. How many journals should be subscribed? How much spent instead on conversion to digital formats? How much on time-consuming human-mediated reference interviews and customized information retrieval? Recovering costs also is a game of relative scarcity or choice: How high a subscription will users pay for a service before they switch to purchase some other service? How much can be recovered from usage and transaction fees before users sufficiently reduce their usage that the mission is not being met?
The library institutions and information systems with which we are familiar were largely developed to structure and mediate solutions to problems in a time when a different set of resources were relatively scarce. A full-text-indexed card catalog wasn't rational, but various full-text searching mechanisms make perfect sense when data storage and processing costs have fallen so far relative to other costs.
Let me offer a concrete example to illustrate my claim that everything is different now: the cost structure of information duplication and distribution. (For an extended discussion of some of these issues, see my recent article (MacKie-Mason and Riveros 1997).) In traditional, print-on-paper publishing, duplication was costly and there was a strong trade-off between duplication and fidelity. (For example, it might be less expensive (ignoring for now any expected punishment costs for copyright violation) to photocopy of a book than to produce an incremental offset-printed copy, but quality suffers.) With a digitally published document, duplication and distribution cost is rapidly approaching $0 with fidelity of 100%. To the extent that the provision of publishing and distributing services is competitive (which they seem to be, at least in the long run), there will be strong pressure to sell services that cost $0 for a price of $0. Zero or low per copy (or per transaction) pricing will not recover the high cost of first-copy production, however. Thus, if we wish to see socially desirable competition and widespread distribution of documents (that cost $0 to distribute), we will need to develop mechanisms for recovering value from users other than through traditional pricing per physical replication. Most of what we know about pricing and product bundling is useless. New work in this area is very preliminary and still several layers abstracted from realistic implementations. In addition to my paper mentioned above, see Bakos and Brynjolfsson, Chuang and Sirbu, and Varian (Bakos and Brynjolfsson 1997; Chuang and Sirbu 1997; Varian 1997).
The previous example assumes that publishers are able to retain control over material they publish, and that the problem is from competition between publishers when incremental duplication and distribution costs are near zero. Of course, zero duplication and distribution costs create another problem: unauthorized production. To even get value-adding publishers to participate we must find means for them to recover sufficient return for their efforts despite the ease of uncontrolled reproduction and distribution. Pam Samuelson has much to say about this problem.
My main point is this: we are here today because there have been radical
changes in the relative scarcity of key information system inputs. Our
research efforts will be rewarding if we focus on how the drastic, and
unusually fast, change in relative scarcities challenges slow-changing
institutions and information allocation mechanisms. I will now briefly
suggest a few priority areas for research.
Likewise, research so far has ignored the time dimension and the allocation
of uncertainty between publishers and consumers. With a pre-paid subscription,
the consumer bears much of the uncertainty about product quality; with
various post-paid access schemes (e.g., document delivery and bundled variants),
the buyer chooses the articles after they are published and quality
is (at least somewhat) resolved. Together with one of my doctoral students
I am doing some research on how the allocation of uncertainty affects the
design of information good bundles.
One can define the "quantity purchased" of information goods in a variety
of ways. At root, there is the notion of a "transaction". The user receives
N articles, or N accesses to an article, or N accesses to a database, &c.
Depending on how a transaction is defined, nonlinear pricing can be isomorphic
to a bundling structure. As far as I know, the relationship between nonlinear
pricing and bundling has not been explored.
Consider, for example, a scholarly journal. The archive of one journal title can be thought of as a bundle of annual volumes. A volume is a bundle of issues. An issue is a bundle of articles, editorials, letters, notes. Each article may be a bundle of a bibliographic header, an abstract, text, figures, equations, video clips (increasingly common!), data files, references, &c. Various bundles could be constructed along any of these dimensions; alternatively (or in addition!), pricing could be nonlinear in the quantity of any or all of these items included in a bundle. Further, goods could be differentiated by various time and duration characteristics: one could purchase access by the hour, day, month, year, or for an unlimited period. Bundling and pricing could also be differentiated in terms of what rights to use are transferred: read only, read and print, archive, redistribute, &c. See figure 1 in MacKie-Mason and Riveros (MacKie-Mason and Riveros 1997).
Traditional publishing and distribution institutions and market forms have evolved with negotiation (pricing and/or bundling) flexible on only a small subset of these many dimensions. Which dimensions are variable was probably determined by the interaction between the heterogeneity of consumer preferences across different dimensions (if everyone wants a bundled journal, why offer to sell by the article?), and the costs of production and distribution (it is much more costly to separately print, bind and distribute by the article rather than by the journal issue). As discussed above, the cost side of this interaction is changing radically, and therefore we should expect to see major innovations in the configurations of bundles and pricing schemes. We should be thinking about how to structure and manage the complexity that is enabled by decreasing reproduction and distribution costs.
Therefore, I think we need research on languages, tools and architectures
for understanding, negotiating, contracting and enforcing contracts on
multidimensional information product spaces. We need common terms and conditions
that are flexible enough to accommodate the vast potential complexity of
the product and price space, without creating such high transactions (including
legal) costs so as to kill the additional value that can be created by
innovative bundling and pricing schemes. Wellman, MacKie-Mason and Jamin
are exploring an architecture for a generalized information resource
as part of our work on an automated information service exchange system
(the MARX project).
Work on a rights management language is being pursued jointly by IBM
and Xerox.
There is more than one conceivable way in which to reward intellectual property creators for their efforts. (I will ignore other policy objectives, such as protecting a creator's reputation from unauthorized transformations of her work.) One broad class consists of property right control, so that the creator can extract compensation for the use of the intellectual property. Without property right control, copies could be made and distributed without compensation to the creator. (A rather different class of compensation methods consist of direct subsidies to creators, paid by a government entity or privately on a work-for-hire basis.)
Since the concern of a property rights control method is to modify individual behavior -- viz., to discourage users from using intellectual property without providing compensation to the creator -- it is convenient to think about control methods as incentive schemes. (When dealing with human behavior, "prohibition" is really an incentive structure: people choose whether to obey the prohibition based on a balancing of considerations.) There are at least two broad types of such incentive schemes: legal incentives, and market or cost-based incentives. A legal approach codifies property rights, along with terms and conditions for the exercise of those rights, and provides an exogenous enforcement mechanism backed by the police power of the state. A potential violator of a property right considers the balance of benefits from use without compensation and the expected costs of legal penalties in order to choose whether to respect the property rights or not. Similarly, a market approach creates costs of use without compensation which must be compared to the benefits of such use. The primary difference between the two approaches, in some sense, is that the legal system uses centralized coercion to impose its costs, while the market system uses distributed implementation of the costs.
Let me make the market approach concrete with an example: software copy protection schemes. Suppose shrink-wrap licenses, &c., do not provide enforceable legal rights on software, but publishers distribute software in forms that are somehow "locked" or copy-protected. As we know well from the early days of the PC industry, copy protection is rarely invincible -- but to get around may be costly. Therefore, a potential user of intellectual property is faced with the choice whether to pay compensation to the creator or to pay the costs of circumventing the copy protection. Another, increasingly popular market-based scheme is to tie together an easily-appropriated information good (e.g., a software program) with some good that is very costly to appropriate (such as customer support or product customization).
Of course, in practice legal and market methods are intertwined: violating copyright by photocopying a book is discouraged both by the expected legal penalties and by the cost of photocopying.
The balance between these two methods of providing compensation to creators depends in large part, I assert, on the relative costs and efficiencies of the methods. The radical decrease in the cost of silicon and sand technologies shifts the cost balance. Sound policy needs to be informed by a consideration of what is now and will soon be possible given the changes in relative costs. For example, it may be that creators will on average be able to obtain higher returns than before simply from market-based methods of recovery; thus it may be that we do not need a costly and possibly clumsy legal apparatus to accomplish the policy purpose (see, e.g., Schlachter (Schlachter 1997)). Conversely, the relative cost shift may be more helpful to those who would appropriate intellectual property without providing compensation, in which case we may need a more restrictive legal method if we wish to maintain the same level of compensation to creators.
Interesting work is being done on rather new forms of compensation recovery.
Examples include the work on trusted
systems by Mark Stefik at Xerox PARC (Stefik 1997), "cryptolope"
type technology such as that developed by IBM, and the ideas of superdistribution
popularized by Brad Cox at George Mason University (Cox 1997).
The collapsing cost of silicon and sand mean that more information than ever is easily and immediately accessible to tens of millions of users. However, the nature of the Internet medium is such that most of the social institutions that evolved over hundreds of years to provide authentication and evaluation are no longer very functional. Even though we were often warned not to believe everything we read, it usually was pretty reasonable to believe that if a public library purchased a hard-bound book, the purported author was in fact the author. And we had a sense that the volume had been evaluated by an authority with at least some reputation and trustworthiness. Today, we often can't count on the authenticity of a digital document, nor are there authenticated authorities with established reputations who can provide evaluations and assurances for most publicly accessible digital information.
Some of the most interesting work I have seen recently concerns the
design of evaluation services, often using the approach of social or collaborative
filtering. Most of the attention has gone to the computer science necessary
to implement such systems (and to a largely orthogonal question of the
role of artificial agents in such systems). I think the more interesting
work is in the area of economics, sociology and interface design: what
will it take to get people to use such systems? How can trust be established?
My colleague Paul Resnick
is doing work in this area, along with others. See, for example, "The
Market for Evaluations" (Avery, Resnick et al. 1996).
Bakos, Y. and E. Brynjolfsson (1997). Bundling Information Goods: Pricing, Profits and Efficiency. The Economics of Digital Information (tentative). B. Kahin and D. Hurley. Cambridge, MA, MIT Press.
Chuang, J. C.-I. and M. A. Sirbu (1997). Network Delivery of Information Goods: Optimal Pricing of Articles and Subscriptions. The Economics of Digital Information (tentative). B. Kahin and D. Hurley. Cambridge, MA, MIT Press.
Cox, B. (1997). "Objects as Property." IEEE Software Magazine(January).
MacKie-Mason, J. K. and A. Jankovich (1997). "Pricing Electronic Access to Knowledge: A Field Experiment." Library Acquisitions: Theory and Practice.
MacKie-Mason, J. K. and J. Riveros (1997). Economics and Electronic Access to Scholarly Information. The Economics of Digital Information (tentative). B. Kahin and D. Hurley. Cambridge, MA, MIT Press.
Schlachter, E. (1997). "The Intellectual Property Renaissance in Cyberspace: Why Copyright Law Could Be Unimportant on the Internet." Berkeley Technology Law Journal 12(1).
Stefik, M. (1997). "Shifting the Possible: How Trusted Systems and Digital Property Rights Challenge Us to Rethink Digital Publishing." Berkeley Technology Law Journal 12(1).
Varian, H. (1996). Differential Pricing and Efficiency. Berkeley, CA, School of Information Management and Systems, University of California.
Varian, H. (1997). Versioning Information Goods. The Economics of
Digital Information (tentative). B. Kahin and D. Hurley. Cambridge,
MA, MIT Press.