| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
A recent article by Mitch Waldrop (Waldrop, 2008) in Scientific American described a trend toward adoption of on-line collaborative tools in science. Waldrop uses the term "Science 2.0" for the trend in the same sense that "Web 2.0" is a catch-all for the now-ubiquitous collection of social networking sites such as Facebook or LinkedIn. A key notion is that these types of tools offer the potential for accelerated scientific discovery based on facilitated collaboration and the sharing of experimental data and methods. The tools range from the preprint servers that have been around in the field of physics for some time to data-mining sites (e.g., http://www.NeuroMorpho.org) where independent laboratories deposit experimental data by using standardized meta-tagging protocols. Central to the recent momentum of Science 2.0 has been the endorsement of collaborative data-sharing by funding agencies (Teeters et al., 2008) and the Open Access movement in the world of scientific publishing (Kaiser, 2004; Harnad, 2007). From a policy perspective, there is an emerging consensus that the tools of Science 2.0 may offer the taxpayer the biggest bang for the buck: however, nagging concerns remain about issues such as fair use, human subject privacy protection, intellectual property, and quality control.
Consider then the dilemma faced by our current and future authors: which parts of Science 2.0 are likely to become the norm? And in turn, which will eventually be consigned to the dustbin of scientific fads? Further, what will be the role of a 100+ year-old trans-disciplinary journal like The Biological Bulletin in the collective scientific consensus that will no doubt emerge? None of these questions strikes me as having an easy response. On the other hand, it is clear to me that for our authors, and for the Journal itself, getting the answers right is critical. This is particularly true in the many areas of biology where massive datasets, especially images, are the engine of scientific progress. Image data is inherently complex, large in scale, and easy to manipulate. Provenance of such data is crucial (Mackenzie-Graham et al., 2008). Issues such as segmentation, feature extraction, and statistical analysis present both opportunities and risks to investigators who adopt Science 2.0 as they might add a new conventional methodology to their armamentarium of tools. When we consider publishing results based on "shared" imagery, whom do we hold responsible for the integrity of the experimental data: the manuscript authors, the curator of the database (or website), or the original "owner" of the data? How would we handle the review process if a reviewer was the original producer of the data but is not an author of the manuscript? Along the same lines, for biomedical journals, how do their editors handle human subject protection when, as a result of the sharing process, the data has been entirely scrubbed of information that would allow compliance with safeguards to be verified?
It seems obvious that our original questions raise further issues, none of which have answers that are obvious.
Let us step back for a moment and approach Science 2.0 from a different direction. Consider the question of what it is to conduct modern biological research. In the spirit of Popper (Popper, 1934), we start with a theory that generates a hypothesis and then an experimental design. We conduct experiments to collect data, we test the experimental data against the hypothesis using statistical tests, and we interpret the results in the context of our hypothesis and either tentatively confirm or evolve our theory. From my perspective as editor of The Biological Bulletin, there are two key aspects of this process that Science 2.0 can potentially change for the better. First in the area of assisted- or automated-hypothesis generation: the terabytes of new data produced yearly in the field of biology make it a challenge to generate hypotheses that are truly parsimonious with the latest data (i.e., the experimental results are evolving more rapidly than our ability to create hypotheses based on them). The collaborative tools of Science 2.0 have the potential to deploy the combined power of multiple scientific minds, enhanced by high-performance computation, to constrain hypothesis generation while at the same time accelerating it. Imagine a virtual space where the posting of raw data online from one laboratory eliminates the need for an experimental control condition in another while at the same time automatically generating the "tweak" in the second laboratory's working hypothesis necessary to accommodate the new information provided by the first. Computational models may enhance this process, always of course constrained by the results gleaned from bench-top experimentation or field observation. If this idea sounds ripe with potential, it is also rife with potential problems, especially quality control.
The second area of the Popperian process that might be enhanced is in the realm of data-sharing. Much of the collection of biological data is expensive both monetarily and from the standpoint of labor (in my own field, the complete thee-dimensional reconstruction of neuronal structures falls into this category). Hence generating the statistical power necessary to reliably test a hypothesis can be a problem, especially at the frontiers, where the ability to even resolve the data depends upon sophisticated instrumentation. Here Science 2.0 can create spaces where raw data from multiple laboratories can be deposited for sharing (increasing the potential statistical power of any analysis) and then analyzed by multiple means in much the same way that large-scale survey research and longitudinal clinical trials are re-used to test multiple hypotheses. Here, in addition to the problem of quality control, is the need for exquisite attention to annotating the raw data. This annotation requirement assumes that different laboratories are using common research platforms. Although such platforms are becoming increasingly deployed (e.g., functional magnetic resonance imaging), they are a long way from being universal. Until they are, subtle differences in methodologies and experimental design may impede the advent of Science 2.0.
Another area of potential difficulty is a result of current promotion and tenure procedures in academic science, where rewards are allocated on the basis of impact and priority of peer-reviewed publication (in addition to success with grants). Thus, investigators have real reasons to fear being "scooped," and these concerns may hinder the type of collaborations envisioned above. The sharing of data, as opposed to conclusions based on that data, prior to publication may ameliorate some of these fears; but in truth, significant progress toward Science 2.0 may await more significant changes in the academy.
If we consider Science 2.0 first and foremost from the above scientific context rather than from the standpoint of professional advancement or science policy, then it seems to me that a way forward emerges, both for the investigator and the Journal. This virtuous path can be summarized with the following precepts:
From these four principles that I propose, a potential author may plan the use of Science 2.0 in a submission and, at the same time, editorial judgment may be rendered.
 |
Literature Cited |
|---|
 TOP Literature Cited |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
