One way to think of the mission of Science Commons, Wilbanks tells us, is to spark generative effects in the scientific world much as we’ve seen them in the online world. He quotes Jonathan Zittrain’s definition of generativity, from “The Future of the Internet… and How to Stop It“: “Generativity is a system’s capacity to produce unanticipated change through unfiltered contributions from broad and varied audiences”. This raises some provocative questions, when applied to the world of science: “What does spam look like in a patent system? What does griefing look like in the world of biological data?”

The truth is that the scientific world is far less generative than the digital space. He proposes three major obstacles to generativity: accessibility, ease of mastery, and tranferability. He points out that, as science has gotten more high tech, it’s far harder to master. The result is hyperspecialization: neuroanatomists don’t talk to neuroinformaticists… “and god help you if you cross species lines.” And so universities are making huge investments to try to encourage collaboration: MIT’s just build a $400 million building – the Cook Center – to force collaboration between cancer researchers… and predictably, researchers are fighting the mandate to move in and work together.

People approach Science Commons based on encountering Wikipedia or free software, and say, We what ‘that’ for science.” Unfortunately, there’s not enough analysis of what makes those projects work. We’d think that science is the perfect space for this sort of peer cooperation, based on Thomas Merton’s observation: “I propose the seeming paradox that in science, private property is established by having its substance freely given to others…” Scientists solve a complex game theoretic problem with new research: they’ve got to disclose to get credit, but as they disclose, they enable competitors in the field. The assumption that science is easier than the cultural space to build a commons in might not be true.

Because John comes from a legal background (and from the Berkman Center and Creative Commons), he tends to think about legal constraints and protections in the field. He asks us to consider three aspects of the world of science: data, tools and texts.

Texts are protected by copyrights, and they’re actually pretty simple. The system is near universal, and as Creative Commons has demonstrated, it’s invertable through legal tools like Creative Commons.

Tools are as broad as ice cores from the Arctic Circle, bones from an archeological dig or stem cells. Contracts between institutions – materials transfer agreements – govern their movement, and patents govern them, especially in the life sciences and the energy field.

Data is protected by secrecy and sui generis protection laws. Fortunately, copyright doesn’t attach to raw data in databases, but there are legal tools we need to unlock other aspects of databases, including structure and compatibility.

While these constraints and protections are relevant, Wilbanks tells us that we also need to deal with “the three I’s”: incentives, infrastructures and institutions. Collectively, they work together to slow down adoption of open policies far more than laws do.

For years, the NIH has had a voluntary public access policy, asking researchers to make their text accessible for free on the web within 12 months. 4% of researchers did. Recently, NIH mandated compliance, and now compliance is over 70% and rising. While this only affects NIH-funded research (hugely important to US biological research, but less relevant in other spaces), institutions like Harvard and MIT are adopting open access publishing policies that mandate this behavior.

“The tools we use to open literature from copyrights don’t work in databases – we need different institutions.” Complicated data that isn’t correctly annotated isn’t helpful. And we need to focus on bottom up resistance, the fact that there’s no incentive or mandate to label or format your data. In the publication and text space, we have university and funder participation. We don’t have any of this in the data space.

The tools space involve “physical objects with physical existence – they are not subject to long tail, everything is free realities” of the digital world. And here the incentives work directly against us: “You don’t get funded by giving away your stem cells – the opposite, you get more funding for writing papers only you can write” because you’re the only guy with access to the tools. And so “the resistance is fractal – it shows up in the same form at high and low levels.” And it represents a huge barrier – he shows us the patent workflow for Telomerase. “Any new field with VC’s involved has a patent explosion underway. It’s not just US – China is outpacing us 8 to 1 in filing patents around clean energy.”

We need to consider the problems that we’ll face with an explosion of data. He quotes Bruce Sterling: “We used to produce data faster than humans could structure it. Now we produce data faster than machines can structure it.” Students are now assigned to develop a microbial portrait of a streetcorner – they swab lampposts and garbage bins and are able to get the organisms sequenced in a weekend. “We need a domain name system for data if web effects come into being.” Until we get to a strong enough web infrastructure, we won’t be able to get these positive effects.

Science Commons works on developing new types of contracts, but Wilbanks tells us that the heart of the work is lobbying people to use them and tracking the extent which they’re used. “Measurements lead to incentives in science – everyone wants to maximize on that metric”. The other major change that’s going to push the field forward is the emergence of new technologies, which may put forward new norms.

“Science is heading (back) towards the garage”. You can buy a gene sequencer on eBay for less than $1000 – which allows you to go from the physical to the digital. And you can get your novel genes sequenced on sites like Mr. Gene for $0.39 per base pair, bringing the digital into the physical. Citizen bioengineers are figuring out how to reprogram E. Coli into novel organisms that can detect arsenic in water. You can download tools from a database at MIT and synthesize your new creations at the site of your choice.

“We’re trying to let the explosion of creativity occur,” much as it did in the online space. “The evil people are going to use these tools anyway. If only the bad guys and the government have access, my money’s on the bad guys.”

“We need to decide whether we’re going to have a PC or Tivo for science,” Wilbanks says in closing. The PC model is uglier, but it’s far more generative and creative and it’s what we want to embrace in the long term.

A partial account of question and answers:

Salil Vadhan wonders how these ideas apply to his field, Computer Science. Wilbanks points out that there’s a spectrum of fields with respect to their openness, roughly spanning from math, physics and computer science (where the fields tend to be extremely open) to chemistry, which tends to be extremely closed. In fields that are highly specialized, the consequence is that it’s much harder to make useful abstractions. In open fields like math, Wilbanks tells us that Hippocratic principles prevent him from getting involved.

Eric Von Hippel wonders about the constraints that the patent system put on generativity. Wilbanks tells us that patents tend to be used like trading cards. Massive, mechanized disclosure is going to change this up. The pharma industry did systematic disclosure as a way to prevent patenting of genomes. “Data as prior art, channeled correctly, makes it harder to get an idiotic patent.”

I asked where Wilbanks would put pressure if he were starting the battle for scientific generativity today, independent on Creative Commons or his legal background. He explains that law is actually a pretty useful place to work from – “we don’t take research money, create text, tools or data,” which means we’re perceived as being fairly neutral. And law is an essential component. But the lever is possibly better placed at the point of funder relationships. Funders can demand that you don’t just release your data – you annotate it. You make sure your tools are accessible. And Creative Commons is a great convener to help funders figure out how to do this, with the legal bits disappearing into the framework.

This turns into a discussion of norms and law – Wilbanks quotes a colleague, who’s got the brilliant observation: “Norms scale far better than the law” –

David Weinberger asks his prefered trick question about the organization of the data we’re talking about releasing. Wilbanks correctly identifies it as “a religious question” and offers “the view of my sect”. That view is that we need standard formats like OWL and RDF, but that we need to let people form their own ontologies and let them fight it out in the marketplace of ideas. He’s suspicious of assertions of people (sometimes people involved with the semantic web) who push forward one particular ontology. This resolution is going to be messy… we need to lower transaction costs to allow people to wire big data sets together.

There are excellent conversations as well about privacy and incentives – check David Weinberger’s blog for what turn out to be very thorough notes, especially on the Q&A.

Pew’s annual State of the News Media report made one pass at answering this question in their 2006 edition. They did an exhaustive study across media of May 11, 2005 and concluded that, of the 14,000 stories posted on Google News that day, only 24 unique “news events” were represented. Here’s the quote: “The level of repetition in the 24-hour news cycle is one of the most striking features one finds in examining a day of news. Google News, for instance, offers consumers access to some 14,000 stories from its front page, yet on this day they were actually accounts of the same 24 news events. On cable, just half of the stories monitored across the 12 hours were new. The concept of news cycle is not really obsolete, and the notion of news 24-7 is something of an exaggeration.”

It’s a striking pair of numbers – 14,000 stories, but only 24 actual news events? – but I suspect it’s a bit deceptive. Grab snapshots of an unmodified Google News page today and you’re only going to get a few dozen story-clusters, each containing hundreds or thousands of similar stories. There are many, many more story clusters accessible within Google News… you’re just not going to get them unless you read beyond the front page or customize your front page to track different topics. The 24 news events is an artifact of the front page model, the decision by Google News to present a certain subset of possible stories during the day. It’s a relevant number, because it means that the average user probably won’t see stories outside of that narrow set, but it’s not a fair commentary on the depth of coverage accessible through the site.

What’s interesting in those numbers is the 14,000/24 ratio, implying 583 versions of each story. (That ratio is probably much higher today, with Google News following more news sources.) Jonathan Stray did a very smart analysis for Nieman Journalism Lab, looking at a universe of 800 stories about the alleged involvement of two Chinese universities in hacking attacks on Google. His findings were striking:

800 stories = 121 non-identical stories = 13 stories with original quotes = 7 fully independent stories

Stray coded the 121 non-identical stories that had been clustered together by Google (the clustering algorithms are good, but not perfect – nine stories were unrelated to the specific case of these two universities) and looked for the appearance of novel quotes, which he considered the “bare minimum” standard for original reporting. (Interesting – it’s the same logic that led Jure Leskovec to track quotes to track media flow in MemeTracker.) Only 13 of the stories contained quotes not taken from another media source’s report.

The essence of Stray’s piece is the question, “What were those other 100 reporters doing?” The answer, unfortunately, is that they were rewriting everyone else’s stories. Given the current shortfalls in American journalism, this seems like an almost criminal waste of time. Jeff Jarvis offered the advice, “Cover what you do best, link to the rest”, and Stray’s finding suggests that many outlines haven’t yet embraced this particular piece of wisdom.

I was more struck by Stray’s closing point – that even the mighty New York Times got the story wrong. Two schools are mentioned in the Times’s report, the well-known Shanghai Jiaotong University and the obscure Lanxiang Vocational School. As the Qilu Evening News reported, Lanxiang Vocational school is a school that primarily teaches motor vehicle repair and certifies operators of earth moving equipment – it’s an extremely unlikely hotbed of hacking activity. (Though it’s possible that freelance, nationalist hackers were based out of Lanxiang’s computer lab, Qilu’s account casts serious doubt on reports that a Ukrainian professor was teaching specific hacking courses… in part because no serious computer training is offered at the school.) Perhaps it’s a bit much to expect the Times – though it does have a Shanghai bureau – to be reading a Chinese language newspaper… but as Stray points out, the story had been generously translated by the indispensible Roland Soong and was available on his prominent English-language site.

I’d love to see Google remove or deprioritize those big numbers that run under every story cluster. Yep, they’re useful for visualizing media attention – Newsmap does a beautiful job of portraying what stories Google knows about in visual form using these cluster numbers. But they give an illusion of abundance, where there’s often scarcity. If we knew here were 13 stories, not 800, on the Chinese universities and the Google hacks, perhaps we’d be demanding more access to original reporting. Maybe we’d ask for translation and inclusion of journalism in other languages in these clusters. Maybe we’d become more acutely aware that – in the case of this particular story – the original reporting was done almost exclusively by large newspapers, entities whose ability to do this reporting is increasingly imperiled.