A New Kind of Science Collaboration

Scientific American is running a major article on Science 2.0, or the use of Web 2.0 applications and techniques by scientists to collaborate and publish in new ways. "Under [the] radically transparent 'open notebook' approach, everything goes online: experimental protocols, successful outcomes, failed attempts, even discussions of papers being prepared for publication... The time stamps on every entry not only establish priority but allow anyone to track the contributions of every person, even in a large collaboration." One project profiled is MIT's OpenWetWare, launched in 2005. The wiki-based project now encompasses more than 6,100 Web pages edited by 3,000 registered users. Last year the NSF awarded OpenWetWare a 5-year grant to "transform the platform into a self-sustaining community independent of its current base at MIT... the grant will also support creation of a generic version of OpenWetWare that other research communities can use." The article also gives air time to Science 2.0 skeptics. "It's so antithetical to the way scientists are trained," one Duke University geneticist said, though he eventually became a convert.

Comments

The Cathedral And The Bazaar, anyone?

[iris-n]

As a scientist, I have to say that this model is utterly beneficial. One of the greatest problems we run when trying to replicate experiments is that the dirty lab details are (intentionally or not) omitted from the fine print articles, making us lose quite a time figuring them out. Obviously it would disappear if such openness became the standard.

Although the idea of making science collaboratively is as old as science itself, it merits having a working model (just don't patent it!) and standing the principle quite out.

Oh and I *hate* this marketing way of naming everything like software versions.

Tradeoffs

[LwPhD]

As a professional academic scientist who does both experimental and computational projects, I think there is a very good argument that for many types of science, this sort of approach will fail miserably, even after the technology to take care of it is completely mature. For example, take a genomics project of moderate complexity and moderately broad interest. Such a project may not be SO important or SO interesting or SO difficult as to require an entire consortium of scientists to complete. However, it may be sufficiently complex that it will require coordinated experiments that will cost into the 10s of thousands of dollars and require more than a man year of work to complete. In such cases, it is almost always best for a single lab to do all experiments (for quality control reasons). If a lab were to complete all experiments at great expensive (for a regular lab), why would they then give up that data immediately for others to work on? Sure, it would be quicker, and more insights would come faster. But to be perfectly honest, this would probably decrease the ability for that lab to promote its members by getting priority with good publications. Currently (at least in genomics) there is no way to reward a scientist through contribution to the community in this way. Now, if a way to award credit for this type of work were to be created that allowed:

• students to apply such work to graduation requirements;
• postdocs to apply the work to faculty job applications;
• junior faculty to apply their contributions to tenure review;

then I think this could be a viable system. However, in academia, this is very unlikely for a very long time. It is amazing and wonderful that journals like PLoS are trending in that direction. And it is even better that MIT is pushing from the University side of the equation. But until Science 2.0 methods are explicitly taken into the incentive system of academic review, this type of approach is a non-starter for expensive, time consuming, experimental science. On the other hand, I could see this sort of approach being very useful for computational science. With much data already freely available, it is usually super quick to get certain types of data analyses done, though quality is frequently questionable. (Go to a journal club on a bioinformatics paper if you want hear academic work seriously shredded.) However, this kind of work responds rapidly to the sort of peer review described in TFA. So, perhaps science could start with the bioinformatics model and figure out how to meaningful track credit in that arena before applying the model to experimental work?