Publishing science on the web

I spoke last week at an event at the British Library about the future of the scientific article. It was a lively event - lots of friendfeed and twitter reactions - and it got me thinking a lot about the way we use publication in science.

In my conversations with research staff and leaders at the BL, I ran across this statement. Publishers frequently claim four functions: registration (when was an idea stated?), certification (is the idea original, has it been "proved" to satisfactory peer review?), dissemination (delivery), and preservation of the record. The journal thus provides for both the claiming of ideas by scientists and for the "memory" of the sciences.

But the Web does a lot of this for us outside of science. It's become easy to write and read, and to use Google as a memory cache. The ability to rapidly find relevant information is part of daily life for us outside of science. But inside of science there is complaint that even within one's own specialized discipline, there is too much to read, too many journals, too little time. This doesn't even begin to include the coming deluge of data wrought by the relentless miniaturization and parallelization of a world where data is generated by robotic lab machinery and captured by tiny, ubiquitous sensors.

Wikis and blogs provide almost costless registration and dissemination of new scientific communication. But resistance to wikis and blogs is a feature of science - Nature's web efforts are yet to make significant revenue despite significant individual use. Is it a matter of certification? Preservation? Cultural aspects related to the way we fund and reward scientists?

Another thought on science communication - science is already a wiki if you look at it a certain way. It's just a really, really inefficient one - the incremental edits are made in papers instead of wikispace, and significant effort is expended to recapitulate the existing knowledge in a paper in order to support the one-to-three new assertions made in any one paper. And the papers are written in a highly specialized form of text that demonstrates the expertise of the writer in the relevant domain, but can form a language barrier to scientists outside the domain understanding the key facts.

In places where the local knowledge is sufficient enough to create falsifiable hypotheses and experiments, the time required to learn the language of others doesn't get rewarded by results - gene sequencing doesn't need a physicist, for example. How can we get to enough technical standards so that this kind of science can be harvested, aggregated, and mashed up by people and machines into a higher level of discipline traversal? Right now the problem is we still think about cross disciplinarity as a function of people choosing to work together. But the internet and the Web give us a different model. What's more cross disciplinary than Google? But the language barrier among scientists is preserved - indeed, made worse - by the lack of knowledge interoperability at the machine level. It's the Tower of Babel made digital. Until we can get past that one, we're going to be stuck doing human speed knowledge construction on machine speed data generation...