Tuesday, May 19, 2009

MIT: Tuesday

Session 5: Mixtures of Openness and IP (Moderator: Katherine Strandburg)

Carliss Baldwin: Drawing the Boundaries of Intellectual Property

Competing social arrangements for innovation. Producer innovation: independent inventor, vertically integrated firms in an oligopoly, strategic alliances for knowledge creation, and modular clusters of forms have grown up along with user/collaborative innovation, all in competition and interaction/hybridity. Hybridity where one provides a platform: Valve Software provides an engine for Counterstrike, a user-created game. Modular cluster of firms around Linux: open source sitting within firms making different pieces of the system. Users might also be able to subcontract with firms to do their work for them. Hybrid social relations require IP modularity.

All innovations are new designs or changes in existing designs, therefore designs are the basis of much (but not all) IP. Design’s structure can be represented as links between design dependencies: if A changes, then D, F, Q, and Z need to be reviewed and might need to be changed. Modules of the design exist in a matrix that can be graphically represented. Different designs with the same functionality can exhibit different structures.

Knowledge is an overlay on the design structure, associated with individual elements or groups of elements. Knowledge “chunks” are given an IP status by their owners. Incoming IP you might be using, outgoing IP you might be generating. IP status is the combination of legal protection chosen by owner and owner’s access policies. The IP status of different chunks can be incompatible—shared can’t be secret. Proprietary licenses may be incompatible with GPL licenses. One example of a violation of IP modularity: licensed-in code was distributed throughout the codebase; license was about to expire, creating a classic holdup problem. Solution: redesign the codebase to modularize so that there were no dependencies between the new platform and the licensed code, changing the transaction situation dramatically and allowing the company to get a new license on favorable terms. And quickly they began to use open source instead of licensed code in that module.

Andrew J. Nelson: The Musician-Engineer: Lessons from Three Eras of Technology Development and IP Management in Stanford’s Music Department

Music department has a center for computer research in music and acoustics. Hundreds of compositions, performances, visiting composers, publications. Also has over 100 patents, 39 industrial affiliates, over $25 million in cumulative tech licensing income, which is unusual for a music department.

How does an academic music department come to engage in leading-edge technology development? How has their status as user-innovators affected their attitudes towards IP?

Frustrated by the limited timbre of the 100 usual orchestra instruments, one guy turned to the computer to compose. He wanted to write a new composition and created frequency modulation synthesis, which was cheap digital synthesis. The tech licensing office thought it was an interesting technology and licensed it to Yamaha in 1975. Other contributions to things like surround sound.

1970-1983: applied for only four patents in this period. Almost all the discourse to outsiders is focused on new compositions. New technologies not incremental advances.

1984: revenues start pouring in from the Yamaha deal, and that changes the attitudes. 1984-1997, 100 patent applications. Discourse centered on new tech and commercial possibilities; established an industrial affiliates program and formalized external relations; Office of Tech Licensing develops a TM program and identifies the music dep’t as the likely source of the next blockbuster tech to replace Cohen-Boyer rDNA—poured millions of dollars into developing new tech. Massive failure: department nearly loses its shirt.

1998-now: the educational entrepreneur. 3 patent applications; full embrace of open source; logo now has the Linux penguin! New approach to monetization: summer programs with experts to learn the open source tech in a hands-on way. Dozens/hundreds of industry reps at $10,000/head—nets more for the department than any IP management they’d done before.

Projects: looking at longitudal diffusion mechanisms and the role of public v. private orientations; multivocality in grant applications—how do you pitch the same project to the Defense Department and the National Endowment for the Arts at the same time? And measuring knowledge flows in network evolution and geographic reach.

Joachim Henkel: Optimizing the Trade-off between ‘Open’ and ‘Proprietary’

Profiting from innovation is commonly assumed to require exclusion mechanisms: patents, secrecy, complementary assets. Shortcomings: “free revealing” isn’t included in standard models; no account is taken of interaction effects. How to integrate free revealing into a study of profiting from innovation?

Different control mechanisms (rights, secrecy, revealing, complementary assets, learning-curve advantages, subsidies) and different appropriation mechanisms (use in own products/processes, exclude others, license others, benefit indirectly from others’ use, intrinsic benefit from innovation). Empirical setting: private branch exchanges in the communications industry. They don’t do much free revealing, but he studied contributions to open standards.

Interviews with industry experts: size of patent portfolio turned out to be important. There are interactions between patents and lead time advantage—if you’re ahead in research you can file for patents others won’t have, and inventing around your patents increases your lead times over other competitors—people who say that patents aren’t that important may be ignoring interactions between patents and lead time. New software to capture interactions. Respondents are asked to tell which company will profit more; companies are said to differ in things like patent portfolio, contributions to open standards, time to market, and sales/service quality.

Geertrui Van Overwalle: Patent-based Collaborative Licensing in Genetics

Empirical question: what is the impact of patents related to genetic diagnostic testing on access to diagnostic testing services? 22 top genetic diseases; looked for relevant patents and blocking effects. 250 relevant patents, 145 active, 267 independent claims.

Findings: 15% of claims were almost impossible to circumvent (blocking), 49% difficult to circumvent, 36% easy to circumvent. Expected that there would be lots of blocking patents on genes, but that was not the case (3%). Instead, there were lots of blocking patents on methods (roughtly 30%). What is the best way to deal with patent thickets? Hypothesis: Formal rules of contract.

All collaborative models in genetics are based on the preexistence of IP rights: open source uses IP as a platform; licensing clearinghouses; patent pools. IP can leverage access. Different models may be appropriate for different types of uses—universities; if owners are both producers and users (v. NPEs, I guess).

Sheryl Winston-Smith: IP Rights and Entrepreneurial Innovation in the Medical Device Industry: David, Goliath, and the Patent Office in Between?

Highly concentrated industry: top 4 companies have 80% of sales and of R&D, but there are a lot of start-ups where competition plays a big role—corporate venture capital is key. Big companies are worried about a competitor acquiring a startup and dominating the market, so they take equity stakes. It matters who the founder is. Entrepreneurial clinicians are able to produce more directly relevant innovations. Intersection between human and financial capital: how does that influence outcomes down the line?

How and when do IP rights align interests between outsiders and incumbents? How would IP be addressed if an outside user modified a device substantially? Open/user and entrepreneurial innovation in medical devices involves significant attention to allocation of IP rights. Sometimes it’s done by contract: paying the clinician; consulting relationships in which the company owns the work. Sometimes the entrepreneur licenses the right to innovate. Billion-dollar settlement for entrepreneur who sued Medtronic: interests are not always aligned. Entrepreneurs need capital to get ideas to market—as users they may have insights about unmet market needs. The corporate venture capitalists need innovations—they want breakthroughs every 3-4 years, and the cycle is shortening.

Entrepreneur can usually get a patent; this provides a basis for negotiation. What is the optimal timing of the approach of the entrepreneur to the corporate VC? Patent protection may be insufficient if they approach the VC too early; the corporation may be able to reverse engineer, build around, or find components elsewhere. Or the entrepreneur may hold on too long and never get to market, or the CVC might get the idea anyway and never reward the entrepreneur. She is researching optimal timing now.

Jeff Furman: Who Benefits from Openness in Science?

Standing on the shoulders of giants requires institutions to preserve knowledge: the loss of the Library at Alexandria set the world back. Have to separate selection effect (how does certain information end up within the institution) from treatment effect (the effect of the institution on how the knowledge is used). Knowledge is not randomly designed, for example university v. firm patents. University patents are cited more broadly than firm patents; we could conclude that universities are better at diffusing knowledge, or just that the sets of problems worked on are different—if firms were working on the same problems and still not cited as much, there’d be better evidence for the “better diffusion” hypothesis.

So, look for reasons that are exogenous to the knowledge itself that shift knowledge between types of institutions and see how the knowledge behaves. Biological resource centers: they collect and offer access to biological organisms for research/commercial development—cell lines, microorganisms, tissue cultures, animal models. Peer-to-peer networks function for research tools, but they can break down for personal or discriminatory reasons. A public deposit may work differently/better. 300+ BRCs around the world; the largest in the US is in Manassas, VA.

So, these cells existed in the P2P network, and then they got transferred to BRCs for exogenous resources. Look at how that affects citations to the foundational work related to the cells. Citations rise after deposit in BRCs, but were flat before the transfer. Moreover, the type of publications generated changes—non-US researchers generate more publications; researchers at non-elite US institutions generate more publications; citations in top journals to papers not in top journals rise, as do citations by elite academics to papers by non-elite academics.


Q: given the costs of preserving BRCs, there’s no long tail—so what do people want to invest in preserving?

Furman: People running these want to preserve everything forever, because they are interested in biological diversity. Costs are relatively low for any one material, but large overall. Federal funding in US has been declining, which has made them pare down the collections that don’t circulate much.

Session 6: Law and Policy

(Moderator: Andrew W. Torrance)

Michael Meurer: Dividing the Spoils: Fair Division

Game theorists investigate: what is the appropriate/expected division of benefits from joint action that produces aggregate benefits in excess of individual action?

Suppose A, B, and C could collaborate: A+B+C or A+B or A+C produce 6 units, whereas B+C or any of the three standing allone produce zero. How to allocate if they all cooperate? Egalitarian = 2 for everyone. Proportional: can’t do it because standalone product is 0. Shapley value: 4 to A, 1 to B and 1 to C. Nucleolus: 6 to A, because A is the key in any solution and neither B nor C is required.

The “contested garment” rule in Talmudic discourse: when A claims ½ of a garment and B claims the whole, how do we divide it up? ½ is uncontested and goes to B, and ½ is contested and is split. The nucleolus is plausible because people actually come to it, and to the Shapley value, without any knowledge of game theory: TVA asked how to allocate the benefits of its dam and came up with the Shapley value, a generalized version of the Aristotelean proportionality rule that is sensitive to concerns about team play. Shapley value and nucleolus tend to minimize group defection, where it’s possible to do so.

Do the axioms for the Shapley value or the nucleolus appeal? They may appeal under different conditions.

Wendy Seltzer: Intermediated User Innovation

Even when we’re engaged in P2P file transfer, we’re actually using intermediaries to accomplish the transfer, and those intermediaries have their own incentives: make money, avoid litigation, avoid costs of mediating disputes. Chillingeffects.org looks at the costs of those misaligned user/intermediary incentives. Example: McCain/Palin campaign dispute with YouTube over DMCA takedown; YT had to comply with the DMCA and wait 10-14 business days, at the height of an election campaign, before restoring a video for which notification had been submitted; YT professed to be unhappy with this and invited McCain to do something about the DMCA after the election. Recently, another spat over Miss California and Perez Hilton—Hilton sent a takedown to the National Organization for Marriage’s spot showing him discussing gay marriage. After NOM protested, YT restored the video without waiting the 10-14 days, on the grounds that this was a clear case of fair use.

Possibilities: new intermediaries like the OTW, with their interests more aligned with users.

Eric von Hippel: Policy Implications of a User-Centered, Open Innovation System

Strong brands, based on TM, are a major source of firm profits: strong brands get a premium, 30-40%. What do you do to create a brand? Repeated impressions/creating links. Communities have link-creating behaviors they engage in for reasons other than creating brand communities: entertaining/interesting—they create brand strength, costlessly. There are many community-owned backpacking logos. Among backpacking communities, 85% have their own logo, 20% have identity products (what I’d call promotional goods). Often, they buy backpacks, and sew their own logos over the commercial brand. Wasn’t an attempt to create a “brand,” but people were hanging out together and creating an association between a logo/brand and an activity.

[I don’t think this behavior is properly defined as costless. The things that people do to signal participation in a community may have a spillover effect on backpacking “brands,” or you might say there’s no marginal cost to creating the brand, but then you’d also have to say there’s no marginal cost to most forms of creating a standard commercial TM (though we might categorize certain forms of advertising as pure brand-building).]

These backpacks are all made in about 5 factories in China and Vietnam, with variations specified by each brand owner.

Survey: which would you prefer to buy, a commercial backpack with logo, or a commercial backpack with community logo? 34% of community members preferred the latter, and it had a brand premium for them over the market price. If it could be sold 1/3 cheaper, 2/3rds of people would buy it instead—which it could afford to do because brand creation was costless.

John Wilbanks: Science Commons

Extending CC approach beyond the CC license: standard contracts tilting towards sharing in human- and machine-readable form. Work with people who already want to share stuff and connect them, rather than (directly) trying to change people’s minds. CC0—a waiver of rights in a database collection. Machine-readability allows people to integrate databases and licensing in articles, so you can click through and find the provenance of data.

Next question: can a commons automate pharmaceutical workflow outside a pharmaco? It can help coordinate legal rights and desired behaviors. About 350 vendors are registered, from people involved in publishing to people who ship the raw materials. The financing crisis has actually been helpful in getting people on board, because there seem to be efficiencies from coordination.

Approached by Nike to share information related to sustainability technology. A virtual clearinghouse. Nike has a patent on water-based adhesives and they gave it away; then another manufacturer took it and advertised that it was the first to convert all its factories to water-based adhesive, without giving Nike credit. Nike didn’t like that.

Designing a one-click public license: not viral. Allows anyone to license, requires attribution, and if your revenues are greater than X you pay a yearly fee of Y. Informal working number, $30 million/$50,000. Another, private license doesn’t allow people to take the license if they’re competing with the patent owner. Science Commons is also taking the opportunity to reconstruct a research exemption as part of both licenses. Communities can execute/sign licenses until they hit the revenue cap.

Jeroen de Jong: Statistical Indicators to Inform Policies for User Innovation

User innovation is everywhere, except in innovation policy or statistics: EU innovation policy from the EC doesn’t acknowledge user innovation; OECD doesn’t provide guidelines for collecting user innovation.

Policy makers want to answer these Qs: (1) Frequency: is it huge? (2) Social welfare implications/spillovers. (3) Are there existing (market) failures? (4) What do you want me to do?

Frequency: there are many user innovators out there. Social welfare implications are positive: users tend to develop different products with new functionality, and in emerging industries; user innovation is marked by knowledge spillovers, free revealing, and innovations are often transferred to producers without compensation.

Market failures: no work yet systematically documents market failures in user innovation. Educated guesses: capabilities—user innovators tend to have technical capabilities, so if they don’t, they won’t be able to innovate. Network failures (communication). Indivisibility (modularity). But more systematic and empirical work is needed on what goes wrong.

Demark has a program for user-driven innovation. Grants for Danish companies—hiring ethnographers to document user needs and support them. Dutch innovation performance contracts: subsidize collaborative innovation at 50%. Innovation by a group of 15-35 small firms, under the supervision of a coordinating organization such as an industrial association. They contract with each other. In practice, most contracts focus on user process innovation—e.g., lightweight design of boats to reduce fuel consumption. However, external sharing/publication is not mandatory.

Statistical indicators are very important. Statistics is a main reason the linear model of innovation still exists today, despite criticism: policymakers have no other data. Without changing the indicators of innovation that policy entities measure, we can’t expect policy change. However, OECD is changing.

My question for Eric: What is the problem you want to address with the branding idea?

Von Hippel: Overcharging via brand premium.

1 comment:

Jayesh Badani said...

Thanks for sharing, quite informative.