Monday, May 18, 2009

MIT, afternoon session

Session 3: Studies of the Effectiveness of IP and Open Approaches to Spurring Innovation (Moderator: Katherine Strandburg)

Pam Samuelson: Empirical Evidence of the Importance of Open Source to Software Entrepreneurs

Survey of high tech entrepreneurs, predominantly software, computer-related hardwar, biotech, and medical device firms. Mailed to 15,000 firms drawn from Dun & Bradstreet and Venture Expert, 1333 responses received. Setting aside defunct firms/returns, response rate 12% for software & IT, 25% for biotech and medical devices. Also got external data about nonrespondents and didn’t find statistically significant differences in firm characteristics, except more companies from the West than the East responded (Berkeley effect). Paper to be published in the Berkeley Tech L.J. Spinoff article: look at software respondents, 708 firms in the sample. One quarter venture-backed, ¾ Dun & Bradstreet. 69% answered by CEOs, 12% by CTOs. 85% were brand new startups, not joint ventures or spinoffs; varied funding sources.

Goals: D&B firms most wanted to remain private, Venture Expert (VX) firms wanted to be acquired/have IPO. Average number of employees: 58. Roughly half engineers.

Do they own/have pending application for patents? Software firms: 35.5% yes. VX firms: 68% yes; D&B firms: 24% yes. Non-software firms: 82% yes. Varies by sector: 90% of internet software companies had patents v. only 21% of VX internet content companies.

Why patent or not? They cited protection from copying; enhancing reputation and increasing likelihood of financing/IPO were significant. Nonpatenting firms cited costs as the most significant factor (27%); perceived nonpatentability also mattered. First mover advantage was more important than patents; for nonsoftware companies, first mover advantage was not as substantially different from patents. Patents were last on the list of things that they thought offered a competitive advantage.

More than 70% of software firms use open source; only 34% of non-software firms do. Some variation by subsector. More than 1/3 use open source as part of business model, VX = 30% and D&B = 35%. Least used in internet content (20%), most in internet software (42%). Software entrepreneurs are less likely to seek patents than some expect; they don’t do so for competitive advantage but for reputation and likelihood of getting financing.

Allesandro Nuovalari: Innovation Without Patents (XVIII-XIX Centuries)

Historical story in textbook: “there was no economic growth before the industrial revolution because there were no patents.” This is not so. A large amount of invention occurred outside patent protection. Patents were not necessary for the Industrial Revolution. Patenting rates across industries in 1851 in Britain and the US were very low, often under 10% depending the industry. The “great” inventors (entries in biographical dictionaries)—1650-1850 in Britain, 40% of the greats never took a patent, and yet are credited with at least one important inventors. A different sample of the greats: 32% never took a patent. And this is biased, because the inventors in such accounts are the romantic inventors who tended to work alone, rather than in groups.

Collective invention: the case of the Cornish engine. When the patent expired, surge in growth of improvements. Collective invention was a critical source of innovation during this period. The Cleveland (UK) iron industry 1850-1875, the English clock and instrument makers, and the Cornish steam engine makers after 1800. Contrary to accounts, these are not uncommon cases and they were not vulnerable and ephemeral—these were foundational technologies for industry. Many other examples of collective innovation: Lyon silk industry, Berkshire paper-making; Western steam-boat; Viennese chairs; Japanese cotton spinning; Norwegian brewing; wind power in Zankstreek.

So what were the available alternative to patents? Secrecy/lead time advantages; prizes and awards; procurement by government/military; patronage.

Andrew W. Torrance: Patents and Regress in the Useful Arts

More on the simulation he discussed earlier: allows open source options, licensing, and many other variations. Sometimes you infringe other people’s patents, or are accused of it. Potential strategies: play it safe (avoid infringement); fast and loose; some people patent a lot and sue a lot, others pursue mixed strategies. 30-minute games. Players appeared to enjoy the game—they loved to win (lots of fist-pumping and insults thrown at other players). They tend to seek patent protection where available. Hypothesis: rational strategy of sowing (incurring costs) then reaping. Players like to license, and they like to sue—they get attached to their patents. Modest use of open source.

Further tests: how different parameters affect the results—ease of patentability, ease of enforceability, patent term, prior art, information costs, number of players, game duration, damages/injunctions. If people come back and play day after day, do the results change? What if representatives of different populations play each other? Is there an optimal set of parameters that makes the patent system behave better than the alternatives? What are the parameters for optimal hybrid regime performance?

Fiona Murray: Role of IP Rights and their Enforcement in Knowledge Accumulation

Knowledge production is step-by-step: outputs of one process are often inputs of the next. Knowledge is often multifaceted: key outputs include information, materials, know-how, methods. Knowledge is non-rival and can be input into many follow-on experiments. Mere production of scientific knowledge doesn’t guarantee its use by follow-on scientists. Institutional arrangements matter: disclosure and access can be facilitated or hindered by law, norm-based, firm-oriented or community-oriented solutions.

Supply side: knowledge inputs are increasingly costly. Demand side: scientific community is increasingly internationalized and growing.

Response: formal institutions for material and information access. And formal institutions for licensing and other IP uses.

Empirical questions: how do you know that IP is producing the effects we see? How can we measure the follow-on innovation we care about? How can we compare the alternatives?

Looked to whether events in science produced new authors and new projects: open access to a particular variety of genetically altered mouse did in fact produce these additional materials. When we compare IP to open systems, we see more diversity in scientific research in the open system. But scientific communities have traditionally been hierarchical and status-driven; when we talk about informal norm-based systems, we have to consider whether that will really result in democracy, or only in other kinds of restrictions on participation?

Jim Bessen: Patents as Property

Patents as compared to mineral rights, land rights, etc. It’s said that patents solve the free rider problem and incentivize innovation, and that’s true of ideal patents, but actual property rights might or might not work. We only seem to talk about ideal patents; if we were discussing tradeable pollution rights, we’d be paying more attention to institutional design. We need more discussion of reality/institutional features in patent.

Property rights work well if they’re well-defined and enforceable, but economists rarely define what that means. Without good definitions of the rights, you can get multiple claimants to the same assets. Poor institutions may allow multiple claimants for an innovation—software patents, where RIM and NTP both claim to own. There may also be a technological mismatch, where the economically useful asset has multiple claimants—the anticommons, where multiple owners have formally defined rights but the rights don’t work in practice.

If we have a single owner with relatively certain rights, that’s the classic ideal case. Single owner with relatively uncertain rights = overuse of patents. Multiple claimants with relatively certain rights = anticommons. Mixed—one claimant with certain rights plus many others with low-probability claims, like patent trolls = ambiguous. This last is a situation of notice failure/underuse of patents—a situation where people ignore rights and then you get disputes and litigation.

Practical questions: eBay: do stronger penalties improve welfare? Are “thickets” a problem if rights are ignored, for example in biomedical research and software? Ignoring patents is a type of behavior that comes out of a sick institution.

Rents from patents compared to the risk of litigation costs: for chemical and pharma firms, the benefits far outweigh the costs. For other firms, in the mid-90s, the relatively even balance disappeared and costs tripled over benefits. This is a notice failure situation. Property rights are being ignored; boundaries are poorly defined.

Q&A: Torrance answered a bunch of questions about his simulation, including how you know how much the thing you might choose to patent is worth; this can apparently be tweaked and can change over time as different things enter the market, but is not yet probabalistic as far as I can tell.

Q: Why don’t we see more cocreation across institutions in the sciences?

Murray: At MIT, 50% of publications are authored with people outside the institution. But there are huge variations in number of coauthors, which may have to do with the structure of credit. We don’t know enough about some of these structures. What gets people to meet up in small conferences or other places where ideas can be shared and realized can be dependent on things like whether they can arrange to travel; Linus Pauling couldn’t because he was under suspicion for being a communist, so he didn’t really get his best chance to figure out the structure of DNA.

Session 4: Norms-Based Systems for Innovation and Sharing (Moderator: Karim Lakhani)

Katherine Strandburg: User Innovator Community Norms and Research Tool and Materials Sharing

Inspired by empirical work by W. Cohen et al. showing that research tool patents are routinely ignored by pretty much everyone, academics and commercial researchers. Problems are more evident with sharing of research materials. Researchers as a competitive user innovator community: they invent tools for their own use, benefit from using tools invented by others, and compete for research results so also benefit from exclusive use of their own tools. Simplified rational choice model is a prisoner’s dilemma for a user community composed of identical members (though of course they’re not actually identical). Cost of sharing innovation turns out to be very important, as well as benefits from exclusivity and benefits from sharing. Under varying circumstances you can get: free revealing no matter what; sharing if there’s a norm of sharing; or no sharing because exclusivity is too attractive. Policy lesson: if sharing is socially beneficial, tweak the parameters.

Application to research tools and materials: three key sets of variables that influence parameter values or enforceability of a sharing norm. (1) Garden variety (only useful for research) v. dual purpose (has a use in research and a use outside research, for example a diagnostic test)—affects the benefits of exclusivity. (2) DIY (a tool you can make in the lab) v. material (something that requires material basis or tacit knowledge)—affects the cost of sharing. (3) Academic scientist v. industry scientist—researchers’ preferences affect the benefits of exclusivity.

To promote materials sharing, we’d want to minimize cost to the inventor. Centralize distribution: centralize repositories, license a commercial provider. Increase returns to sharing by giving co-authorship/acknowledgement. Increase penalties for noncompliance with norms: journals or funding agencies can make sharing a requirement for publication or for getting materials. Reputational penalties might also be used. Norm entrepreneurship: to stabilize sharing norms, for example 2007 university white paper on tech transfer encouraging academics to resist onerous licenses and reserve sharing rights within the academic community.

Chris Sprigman: New Research on IP Norms

IP’s negative space. Began research on the fashion industry; recent work on stand-up comedians. Fashion works in a low-IP equilibrium—there’s a mix of sharing and property; TM serves important functions. Stand-up comedy produces virtually no litigation, though joke-stealing is known. Comedians use anti-appropriation norms instead, monitoring by the community.

Coming at the question from the other side: doing experiments on how creators value their IP. Questions: are created goods subject to an endowment effect? Is there some other effect that causes owners of created goods to depart from rational valuation? What are the implications for a rational choice model?

Subjects are asked to write a haiku. Most UVa students can do this. Induced value: $100 at stake. Ten people per section = expected value $10. Ask the students what they’d accept to transfer the chance of winning the prize to a bidder. And ask other students what they’d pay to get the chance of winning the prize. Different treatments: First, authors write poems for a contest, chosen on merit. Second, authors write poems which are lottery tickets: thus they can investigate role of perceptions of quality. Owners are given poems written by authors, and they are subject to the contest/lottery conditions too. Do authors have a bigger gap between willingness to accept and willingness to pay than owners? This is a question about intermediaries.

Maybe no one thinks they’re a below-average author, or maybe they’re worried about their quality and thus underestimate their chances. The experimenters debrief them afterwards to make sure they understood the conditions. Current finding: very large gap between willingness to accept and willingness to pay in the author conditions.

Andrew King: Kitchen Confidential? Information Transfer among High-end Restaurants

His research focuses on decentralized and voluntary institutions: codes of conduct, best practices, sharing organizations—as opposed to involuntary and/or centralized institutions, including culture (involuntary decentralized), government regulation (involuntary centralized), and firms (voluntary centralized). Main finding: self-regulatory institutions arose in response to exchange problems. They arose in the hope that they’d work without a lot of transparency or teeth. But without transparency and sanctions, you get adverse selection/moral hazard; firms get worse by participating. But self-regulation comes into existence when there’s an industry problem, like Bhopal disaster.

Can we expand this to public goods? He and his co-investigators looked at fashion. There are not enough fashion companies to do large-scale confirmatory research, so he’s now working on restaurants, interviewing highly-ranked chefs in Milan and Boston. Main findings: more likely to transfer information to a nearby chef with a similar quality point—similar d├ęcor level, similar cuisine. That is, their closest competitors. That seems absurd. Chefs at a certain level say that chefs don’t copy. But they actually do copy. They just copy at a distance—only people from other places and other career points (students in competition).

Chefs also say they wouldn’t screw their neighbors because they never know when they might need five pounds of swordfish or a bus boy; they expect reciprocity beyond recipes. It’s a repeated game. (Chefs are like farmers and ranchers in Shasta County!)

Research is ongoing. Real problem is separating out monopoly competition from the effect of norms. You don’t want to look like the next guy because you need to maintain product distance, so you wouldn’t copy your neighbor’s recipe exactly even if s/he told you exactly how to make it. How do we separate that from institutional restraints on copying?

Stefan Bechtold : TV Show Formats: A Global Licensing Market Outside IP?

American Idol has an Iraq version, and is in 40 other countries. Big Brother is in the Philippines, and over 60 other countries. Who Wants to be a Millionaire? is in over 100 countries. Farmer Wants a Wife is in Switzerland and over 16 countries. 3-4 large players worldwide develop these formats. Many originate in Europe. Meta formats: I Survived a Japanese Game Show, on ABC.

Name can be protected by TM. But is there any other protection? Copyright is very messy: idea/expression dichotomy would usually deny format protection, though some countries have applied copyright. Limited caselaw concerns the secondary licensing market (where the format has aired in another country)—it’s usually about the primary licensing market, where a producer offers the format to a TV station that just takes the format without paying. Some countries might apply unfair competition law. Trade secret seems unlikely; no known business method patents. Research continues on this.

You’d expect widespread copying. And there is some of that. But there’s also a $16 billion/year licensing market. What are people licensing? He is talking to industry members. Is there an interaction between national IP laws and the national TV show format industry? Are there private substitutes? There is an international format registry with an alternative dispute resolution procedure; a format bible, with standardized elements so you can define your format precisely for later comparison; trade shows in Cannes and Las Vegas.

Hypotheses: (1) social norms in close-knit community, with about 100 firms; (2) uncertainty/licensing not to be sued; (3) licensing tacit knowledge; (4) theory of the firm.

Rebecca Tushnet: Transformative Works/Transformative Workers

Let me restate Eric’s introductory description of his research: lead users innovate where no market yet exists, then collaborative user communities improve and filter innovations. Then manufacturers enter, usually nonincumbent manufacturers coming out of the user communities. Eventually incumbents decide that the market is big and secure enough to enter.

This is actually one coherent way to tell the story of what we now call “user-generated content.” First there were mailing lists, usually run from universities by students or employees, and Usenet newsgroups; later individual websites and archives, sometimes sourced from mailing lists, hosted by individuals. Companies emerged to host personal sites in return for payment or for advertising—Geocities, which is about to shut down. Specific to media fandom, where people create works based on existing popular texts like Star Trek or Harry Potter, a few people started hosting archives for money, at least for a return sufficient to offset costs, by running ads—fanfiction.net is the biggest example in the media fan community. Google bought YouTube and other mainstream sites arrived—again, from the media fan space, FanLib, a Hollywood startup designed to monetize fan contributions on behalf of major media investors. Notably, it went out of business after about a year.

And this is part of the problem that content industries are facing: there really doesn’t seem to be as much concentrated money in user-generated content as there was in traditional media, even if there is as much or more value. Copyright industries have not figured out how to make the kinds of quality improvements von Hippel identifies as allowing manufacturers to profit from user-generated innovations.

Relatedly, there is a problem of digital sharecropping: because so much of the value generated here is affective, and doesn’t directly help the individual creators materially, the issue of exploitation of the workers producing the creative stuff is quite salient. Wendy Gordon most explicitly put this on the table: reciprocity in gift relations may, and quite possibly should, involve material support. Exploitation of creative labor for the benefit of other parties is especially of concern because media fans are predominantly women, whose work has often been expected to be unpaid, naturally. Traditionally, women’s work has not been thought to require incentives, which has contributed to the material inequalities women face.

Against this background: OTW, formed as a nonprofit, with a mission to “own the servers” and engage in public advocacy on issues of concern to media fans. We could use existing archive software to store, categorize and serve fanworks on those servers, but on philosophical and practical grounds decided to code our own. Philosophically: develop programming skills among mostly female fans; this is empowering and may provide possible material benefits outside the fan context. Practically: from long experience with fan archives, it is clear that well-documented code written with others is more sustainable than single-fan projects. If you code it yourself, you take shortcuts and just bang on it until it works, and then if you leave fandom the thing you made may just sit there until it breaks. A multicontributor project is one that new people can take over running without having that same tacit knowledge.

The OTW’s Archive of Our Own and another, unrelated media fan-friendly project, Dreamwidth (a journaling service), have become significant locations for women in open source. Statistics are hard to come by, but everyone agrees the numbers of women in open source are very small; 1.5% of contributors is a not uncommon statistic. DrupalChix say that Drupal has 10% women on the project, which is an order of magnitude improvement that gets them up to “awfully unrepresented.”

A quote from the Archive of Our Own, statistics from last year: “1134 revisions have been deployed to the Beta Archive to date, and we have had five major releases and innumerable small ones. 150 volunteers have worked on [Accessibility, Design &Technology]/Code/Test, many of whom we have trained ourselves in Ruby and other languages; we aim to teach and mentor all, women especially, who want to learn.” That number is now up to 250, 100% female, 21 contributing code, 80,000 lines of code. Women also make up the vast majority of OTW’s systems administrators.

Dreamwidth, a fork of the Livejournal code, has approximately 100 project contributors, 34 contributing code, 75% identified as female. 280,000 lines of code, though shedding lines as it moves away from the Livejournal base.

Why these projects? (1) Goals are of interest to women: the output is something they want to use. (2) Officially woman-friendly. (3) Offer lots of training and opportunities to contribute: you don’t have to be an expert already to help.

Complicated lessons: (1) Commercial and noncommercial domains can’t easily be separated. (2) Power disparities offline matter in what gets done in open source. (3) Semiotic democracy: making culture and making stuff turn out to be linked—people who think they can create one kind of thing are often willing to believe that they can create another kind of thing.

Q&A

Terry Fisher: Where do norms come from? Are they hardwired? Are they adaptive responses to constraints? Are they narrow economic self-interest? Are they culture-specific?

Sprigman: In standup, norms about copying change because the mode of comedy changes—standup becomes more personal and anticopying norms emerge. It’s hard to draw a causation arrow. Is this good? Depends on your priors.

Jonathan Barnett: distinguish between end user innovation communities and intermediate user/producer innovation communities. Much more potential for normative concerns with intermediate user/producers. The fashion industry used to behave like the description of the TV format industry: a design registry, design policing. Used a group boycott against defiant department stores, which got them in antitrust trouble and shut them down. “Haute couture” came from a guild of French high fashion houses with extremely elaborate requirements down to the number of seamstresses one must employ.

Bechtold: TV formats are very commercial; the social norms are generated because of an economic problem for the industry itself. Are there entry barriers caused by this? It started 4-5 years ago, so it’s emergent, and inherently international, which raises another set of issues.

Me: but we heard earlier that every output is someone else’s input: or apparently we don’t really believe that, if we’re making this distinction between end users and intermediate users. My people use the TV format's output as input for their creative works.

Strandburg: One way to look at “ignore patents” as a norm is to deal with dual-purpose tools. If someone uses a diagnostic test in research, no enforcement, but if someone uses it commercially, enforcement. That’s how you go from no-patent to ignore-patents as a rule. Is that a good idea? Recent lawsuit over breast cancer gene patents: the ignore-patent norm serves researchers well, but does it serve everybody well?

Sheryl Winston-Smith: Does the patent/IP system devolve into a winner-take-all system with lots of people unable to appropriate value? What kind of royalties do the creators of the TV format get, or do they alienate their rights for a one-time payment?

Bechtold: there may be no one creator. They have teams that meet to think of new formats.

Brett Frischmann: It seems so hard to define the boundaries of a TV format! Licensing looks so cartel-like; one wonders why it functions.

Andrew Nelson: Highlights the question of what we mean by norms—is norm-guided behavior distinct from rational behavior? In Jon Elster’s view, people who decide to go along with norms from fear of sanctions aren’t engaged in norm-driven behavior; if you consciously deliberate about whether to share, you might not see an immediate change in behavior but that might result in a collapse of a sharing norm.

King: On institutional entrepreneurs—if you look at environmental issues, the original sponsors often ended up doing something different than what they intended. Adopters of best practices used it as a signalling mechanism, not an improvement mechanism, because the adopters were already up to spec.

Jeroen de Jong: TV format—distinguish commercial from publicly funded broadcasters. Publicly funded may not copy as much. [There are multiple versions of Sesame Street, as I recall.]

Wendy Gordon: Chefs leave out crucial steps in cookbooks.

King: True. Causes a problem when chefs go on tour!

Gordon: Do the people in Sprigman’s experiment get to keep authorship credit?

Sprigman: Nothing transfers but the right to get the money.

Gordon: Does that test for what you want to test?

Sprigman: Q is whether their affective valuation of the poem leaks into their valuation of the revenue stream attributable to the poem. We thought about using digital photos instead of poems, but the IRB made that a nightmare.

Gordon: Formats—in the US, you could use trade dress, idea protection, copyright in compilation as means to protect them. Point is not to say that there’s an easy answer, but reminds us that categories that are not explicitly well protected by a particular type of IP are often more protected because expansionist judges have stretched various categories of IP to cover them.

Samuelson: Contest formats are pretty clearly not protected in the US, though.

Jeroen de Jong: the original format, closest to the original’s trade dress, is usually the most successful, so filing off the serial numbers might not be a good idea.

My questions for Sprigman: supply/demand in distribution: there are lots more aspiring screenwriters than movies/TV shows to employ them. Do you plan to test conditions in which success is really unlikely? Relatedly: role of nonmonetary incentives, which have been stripped from the experiment: what would happen if the winner was going to get published in the student newspaper? What would happen if you offered a bargain: we’ll pay you X to publish your poem, or you can have this lottery ticket for $10? My guess: typical X will be well below $10.

A: Yes, he’s planning to test variations. Tell people that U Va. wants photos for a calendar it’s going to put out. If the university says it’s paying v. not paying, what happens? Need to do this with people who aren’t tied by affection to the university, so may have to do it on the web. [I think that you’d also get interesting results with people tied by affection; in fact, that is where some really difficult issues of valuation/exploitation might pop up.]

No comments: