Sunday, May 11, 2008

STS and IP: Innovation networks

Session 6: Papers by Dana Wang (“Constructing and Objectivating Information: The Identification, Sourcing and Meaning of Information in Reference Lists Found in Patents”) and Katherine Strandburg (“Users as Innovators: Implications for Patent Doctrine”)

Wang: Studies the meanings of manmade objects. Individuals and groups interpret the same object differently (as do creators and engagors). Previous studies have looked at misalignment in meaning. When a discovery is adopted by a separate group of investigators, it may become scientific fact. Misalignment is not necessarily permanent. Engagors may form groups that communicate with creators, who come to identify what they need to modify in future products, and creators come to the engagors—what she calls closure. But not many objects are the result of iterative product development.

Example: patent reference lists. Does cross-citation signal something about deference? (Gillespie’s “respect.”) Backward citations arguably reflect inventor’s knowledge or search at the point of invention. Forward citations suggest some quality—value, importance. But few studies have looked at the creation of reference lists.

She conducted a number of interviews and informal discussions; made observations; and drew on archival data. Info in issued patent doesn’t reflect well the process by which a patent is created. In the file wrapper, there may be a bit more—e.g., the names of all the examiners, not just the last one. But there’s still a lot missing.

Why reference lists are created? Some social scientists have treated reference lists on patents like citations in publication. Not so! People disclose to comply with institutional roles and procedures: the applicant’s duty of disclosure, candor and good faith; the examiner’s compliance with examining procedures.

The information on an issued patent comes from the patent counsel/agent and the patent examiner. The examiner adds knowledge from his own research/expertise and the applicant. He adds references, and may use them in an office action. The counsel submits references from her own search, including foreign patent search reports, and the inventor. There are other related and unrelated sources—paralegals, third parties, search firms. Some of that information gets purged on the way through the application.

The source of the info isn’t important to the ultimate patent; the question is how the info is used to limit the patent claims. In fact, only a fraction of the reference list is discussed in the body of the patent or in an office action. Some information that’s considered general and not material at all gets disclosed. Likewise, when the patent issues, only a fraction of the participants appear on the patent.

Among the issues—a patent attorney may not go through 500 references of dubious relevance if the client isn’t willing to pay.

Strandburg: Users and the role they play in invention—what’s the impact on patent doctrine? Patent doctrine assumes there’s a seller-innovator, motivated by the prospect of commercial sale. Users just consumer. But, as Eric von Hippel’s work shows, the relation between tech and users is more complicated. STS reminds us that tech configures users, and users also configure tech.

Digital mechanisms have further enabled user innovation, but it’s always gone on. Manufacturing processes, computer programs, sports moves, medical procedures, research tools—the list goes on. We ought to think of business methods as user innovations too. It can be either commercial or noncommercial, but it’s not motivated by selling the invention.

Thus, if we think that the patent system has something to do with incentives, we should care about different motivations. Von Hippel’s work shows that users tend to make different kinds of inventions—users make cutting-edge functional improvements relating to very heterogenous needs, where manufacturers make innovations targeting the average users—convenience, safety, using sticky info about mass markets, taking advantage of returns to scale. There are good economic reasons for this; users have sticky info about their own needs, and they get both extrinsic and intrinsic rewards (it’s fun, it increases your human capital) from inventing.

Are there ways for patent law to separate out user innovations, where patent’s incentives are less important? Prior work: We might want to limit the scope of patent law, and include a use exception for infringement.

User innovation tends to happen in communities. Reputational incentives help make user innovation work. If you disclose freely, other users will improve your work and share their results. Sometimes users aren’t in direct competition; even when they are competitors, they may benefit from some shared knowledge. E.g., if they’re in competition with outsiders—ironmakers shared information, but were in competition with other makers.

She made a model of a competitive user innovator community composed of identical members. Components: value of shared invention to each member; number of members; cost of sharing; benefit of exclusive information; benefit of sharing. There is a collective action problem because sharing would often be optimal but people still won’t do it. Often the cost of sharing is low because people are good at picking up on useful innovations—users are highly absorptive. A sharing norm can solve this collective action problem.

She applied this to the sharing of research tools within a scientific community. Scientists doing research are user-innovators, and they’re competitive because tools allow more publication (and grants). Will they share? There’s a relatively large value in using others’ tools because they are useful, and academic researchers like to learn and move the scientific enterprise forward. The ability to keep info exclusive is limited, because grad students graduate and move. The cost of sharing is low, because you just have to publish it. The rewards are relatively high—you get credit for publication; reputation; the ability to place grad students well; possibly collaboration. So a sharing norm is beneficial. And in a tight-knit community, a sanctioning norm will work.

Research on scientific preferences: (1) performing research; (2) autonomy in research are primary. The community controls resources like grants through peer review, so sanctioning mechanisms are available.

What about patenting? That can allow you to publish and not share. Even if your grad students move, they can’t take your invention with them. But sharing will still often be optimal because of the utility of tools to others. Will there be an anti-patenting norm? It was for quite a while.


But increasingly there’s an overlap between academic and commercial research. The value of exclusivity may be greater and the norm may destabilize. She argues that if you can distinguish commercial and research use then the norms can do so as well—ignore patents used for research, which can be extended to commercial research as long as researchers aren’t sell. The evidence suggests a norm of ignoring patents just like this. Wes Cohen’s work also shows that sharing tools is common, but there are greater problems with materials transfer, as to which there is currently not much of a sharing norm. It’s easier to keep materials secret and costlier to share them; materials may be difficult to duplicate.

How to encourage a materials sharing norm? The expanding size of the community itself increases the cost of sharing. Centralizing distribution/diffusion of tacit knowledge might help.

What happens in a nonidentical community of researchers—scientific v. industrial researchers? What of norms then? A mixed sharing norm; norms to police the boundary—everyone reports more trouble getting materials passed across the academic/industry interface. There are some norm entrepreneurs: university white paper on licensing IP for tech transfer.

What is the proper view of the public? Possibilities: Potential inventors needing protection—a democratic but seller-centered model; inventors as experts, romantic inventors, needing special protection; public as consumers whose needs should be met/created/taken into account; public as user inventors, needing protection from the overbearing corporate IP community.

Dan Burk: On Strandburg: Contextualizing IP and accounting for specific preferences is a great way to go beyond blunt economic analysis. Also analyzing the idea of scientific norms as they’re operationalized; the myth and the reality may differ. The norms of science are really appealing (see Rebecca Eisenberg’s work) but there hasn’t been much progress since Eisenberg brought the idea to law. Although this is a step away from Chicago-school analysis, it’s not a big step.

On Wang: It’s a cautionary tale on doing empirical research: if you do citation analysis on patents, they might not mean what you think they mean. (It’s just an object. It doesn’t mean what you think.) We really need anthropological investigation of the patent attorney tribe.

Mark Lemley: He agrees with both papers and would like to complicate them. On Strandburg: He thinks she defines user innovation artificially narrowly—internal uses for commercial or noncommercial purposes, but there are also users who innovate and then resell the innovation. In copyright, innovators aren’t just people who modify works for internal purposes, but also and mostly people modify works for further dissemination. Perhaps this limitation on definition is done to make the story work—the user-innovators are less competitive with the original manufacturer/creator—but it’s worth thinking about these other user innovators.

Also, the role of agency costs/agency problems in institutions. “Companies” and “university professors” are not autonomous decisionmakers. The university professors—who may have a foot in both camps—are conflicted about patents, and their universities are profit maximizers when it comes to patents, and their relationships with profs can be complex. Similarly, the profit-maximizing company is made up of employees who may value openness and sharing, which has driven the production of open-source software even by for-profit companies.

For Wang: This is persuasive and probably conclusive evidence that people who look at citation counts should be doing something else. Legal empiricists haven’t done much of this, but social scientists from outside law shouldn’t be using citation counts to measure innovation or anything else. You can layer onto her case studies a number of reasons to believe that there’s systematic nonrepresentativeness. There are massive industry-specific differences in behavior, including search behavior and in erring towards underdisclosure or overdisclosure (pharmaceuticals overdisclose). There are also likely to be big variations by company/inventor/patent attorney. There is also evidence of strong differences between patent examiners; for the first two years, they work hard, and then they get promoted and stop doing so much work. Can transparency solve this problem? He thinks pretty clearly not. But making clear where info comes from is worth talking about.

It’s not just citation references that are socially constructed in the patent document. The patent claim is the crown jewel of the socially constructed item. It is bizarre that what controls in patent law is the patent lawyer’s words, modified by the patent examiner, then glossed over in claim construction by litigators and judges, with no connection to the inventor at all.

Torrance: When he first worked as a patent attorney, he spent 30 hours looking at materiality of references and got a huge slap on the wrist. He was told to spend no more than .2 hours on materiality, and ideally give it to a secretary/paralegal—it’s pro forma. The manual says materiality is important; practice said it was something to avoid billing.

Another question: what about peer-to-patent? Open source patent review.

Wang: The current system allows it; she saw it once in 35 years, 4000 patent applications, but a third party can disclose through the patent attorney. The client eventually decided to send the letter from the third party on to the patent attorney. It didn’t affect the issued claims, though it shows up in the patent file. Third-party disclosure is not a great way to affect patents. In order to do this, you need large firms with time on their hands reading patent applications who want to block competitive patents, not inventors.

Lemley: The lack of effect is why people don’t do it, though there was clearly strategic behavior on the part of the third party. Also, until 8 years ago, you effectively couldn’t do it because the applications were secret.

There was extended discussion of the truth and utility—two separate issues—of scientists’ beliefs about their own norms. Robert Merton’s work describing how scientists think of their own endeavors was utopian—the line between the descriptive and the aspirational is important and contested.


Cohen: Along with Eisenberg, Arti Rai’s paper asks the same questions about Mertonian ideals in science after Bayh-Dole.

Cohen has concerns about trying to work from the top down by calling these practices “what users do,” when it may be “what humans do.” She compares it to recent work in copyright, where there’s a lot of emphasis on unpredictable individual creation. Rather than thinking about “homo scientificus,” perhaps “homo ludens” is the right model. The everyday habitual practices of users are not necessarily coterminous with norms and what the community thinks it’s doing.

Strandburg: This is part of an overall investigation of different case studies. She wants to ask who the actors are and what they want.

Frischmann: One question: is each generation of scientists still being taught the same norms? This question needs to be asked all the time—Bayh-Dole is just an obvious triggering event.

Barley: We need much thicker descriptions. Open source is not skateboarding; you need to know who skateboarders are to figure out why they do what they do. Merton wanted to describe the particular circumstances of scientists; von Hippel is good on open source but he doesn’t necessarily know why skateboarders do the sharing they do.

Lemley: True, but all advantages are comparative. The perfect shouldn’t be the enemy of the good. Knowing something about skateboarding innovation is better than knowing nothing about it.

Eschenfeld: We build these systems based on assumptions about sharing, and lots of them fail. We may need to know more before we can expect our interventions to do what we want them to do.

No comments:

Post a Comment