Friday, April 20, 2007

Please Step AWAY from the Dark Energy

An interesting essay by Simon White showed up on astro-ph Wednesday night entitled "Fundamentalist physics: why Dark Energy is bad for Astronomy." The whole thing is an interesting sociological read, highlighting in very clear terms the adage (mantra?) that astronomy doesn't want to become like particle physics:

Dark Energy is the Pied Piper's pipe, luring astronomers away from their home territory to follow high-energy physicists down the path to professional extinction.
This topic is big enough that both Rob and Sean have already commented on it (I was travelling! no interenet!). As someone who got into astronomy because cosmology was so interesting (I didn't like astronomy as a kid, and MIT doesn't have a separate astronomy department), I find this topic to be quite enlightening. Sure, we could all be working on what is arguably the "most interesting" topic, but what's the fun in that?

What I found most interesting in the essay, though, were the cold hard numbers behind the reality of the situation, i.e., how the culture of writing and citing papers has changed in the last, say, 30 years. The final figure in the essay outlines how the number of refereed papers, distinct authors, authors per paper, and references per paper have changed from 1975 to 2006.

The numbers in parentheses are the 1975 values.
By 2006 the number of authors had quadrupled but the number of papers had only doubled. On the other hand, the mean number of authors per paper also doubled, so that the number of papers signed by a typical astronomer remained constant at about 2 per year. The size of the astronomical community has thus increased dramatically and a drop in the mean productivity of its members has been masked by the tendency for more individuals to sign each paper. In 1975 over 40% of all papers in the major journals had a single author and fewer than 3% had 6 or more authors. In 2006 only 9% of papers had a single author while almost 28% had 6 or more authors.
As an extreme example, the fourth ranked astrophysicist by citations to papers published over the last decade has never written a first-author paper for a refereed journal and has gained almost all his citations through his right to sign official papers by a large collaboration in which he played a purely functional role
Why these trends occur and just how bad (or good) they are is harder to pin down. I know that big long author lists are extremely uncommon in most fields; are there even any other fields within physics that support papers with more than 6 authors on an "oh, that's normal" basis? I doubt we've found a nice equlibrium within the collaboration system yet: the benefits of "more people means more ideas" has to be balanced out by the fact inidividuals can get lost, e.g., that after the first author, it's difficult to track down just how much each author contributed.


Fredrick William High said...

chris stubbs my adviser wrote a letter to nature on the subject:

i believe it's nature that requires co-authors to list what function they each performed toward the result.

look, long author lists are what happen when your experiments move from the tabletop to giant underground accelerators and large synoptic sky surveys! as such, we need to find a better measure of our quality of work than the length of the publication list.

Vincent said...

The number of references in each paper naturally grows with time as a subfield of astronomy matures, or as a particular object of interest receives ever increasing study. Astronomy, like most sciences, advances by building on an ever-growing base of knowledge provided by the work of others. Perhaps present-day astronomy is more "mature" on average than it was in 1975. In radio astronomy, for instance, if you studied an object or a molecular transition in 1975, there wasn't that much work out there for you to cite. Totally new objects and lines are much rarer today.

Other possibilities abound. With an increased emphasis on multiwavelength studies, the range of work relevant to be cited is increased. Perhaps today's astronomers are also more aware of their responsibility to give proper credit where it is due and are therefore more diligent in referencing. The existence of sources such as ADS and astro-ph also make it easier to find relevant references in the first place.

Nick said...

Individuals do get lost. Here is the LIGO Scientific Collaboration author list for our current science run, S5. Can you find me, mollishka?

Regarding some collaboration paper, I have been asked, "Which six words did you write?" That made me smile.

I do think, though, that the move to big science is a somewhat natural progression. If you want a bigger, shinier instrument, you bring together some like-minded colleagues and pool your pennies. However, entering into this non-traditional family unit can lead to some culture shock.

Fortunately, the particle physics people have raised awareness and acceptance of the issue. I am under the impression that over the past few decades, they have seeded the concept in most physicists that first authorship is often not possible in a large collaboration and that an individual shouldn't be penalized for it.

mollishka said...

Nick: through the powers of alphabetization, you're pretty easy to find on there. But that is an impressive 3-page list.

We actually discussed this "screed" in Coffee yesterday, and one point which was made which is relevant to e.g., LIGO or SDSS, was that in the "particle physics model" it's very difficult for a younger person to find a niche and make a name for themselves. With SDSS, it's possible for a small group of people to use the big data set to do some investigation, and then instead of everyone and their brother being on the author list, they can cite the papers talking about the telescope and data-reduction pipelines and so on. As an outsider, LIGO seems to be more like a particle physics experiment. What do you think? You say individuals get lost, but do you feel people within the LIGO collaboration actually know who you are, and do you feel like you are able to be creative and make an impact?

LSST was also brought up as an example of an "experiment." I think part of the point is to ensure that not all of the data is thrown away (which particule physicsists are exceptionally good at because they have to be); in astronomy, one person's background is another's data.

And then a lot of the points Vincent makes about different factors affecting author/paper/citation/reference ratios is one reason it's difficult to untangle just cause and effects, good or bad, are.

Nick said...

I do think that an individual can get notoriety, fame, and/or respect in a large collaboration. The go-getters emerge as go-getters and get jobs. And, as in any context, good advisers don't leave their students behind. I feel very fortunate that my advisers have given me high priority, high visibility work and made sure that I am giving the presentations inside and outside the collaboration. If anything, I'm more visible than most of my peers, simply by having so many (it's really a lot) of collaboration meetings. Inside the collaboration, 200-400 people know my name and my work because they hear it every few months, even if they checking email at the time. Outside of the collaboration, my group has determined that allocating money for frequent travel is a priority, so I appear in public more often than the average grad student.

The high visibility in front of the collaboration, plus my having developed specialized skills that my collaboration values, I feel very confident that I will have a job when I emerge from the grad school cocoon. Outside of a large collaboration, not many can say that.

mollishka said...

Now you've got me thinking about other kinds of measures of how much a person has done.

There's the h-index, where h = {the number of papers a person has authored with h or more citations}. ADS makes figuring h out very easy; mine is 3 if you include non-first-author papers and 2 if you don't.

Then there are citation counts v. normalized citations; I have 21 citations, but a total normalized citation count of 2.66 (thank you, STEP2). The ratio between these is 7.9; I'm not sure what that tells us though. A measure of how much is actually gained per paper per author, perhaps? We could always just look at the average number of authors per paper, regardless of citation count. Mine's 47 (not distinct, of course) of the one's that have been cited; it jumps to 75 if we include everything ADS has on me because when writing the supernovae telegrams we felt like putting everyone on the mountain on the author lists.

In conclusion, numbers are fun.

Dr. Lemming said...

I think that figure would be more informative if it also included average grant money consumed per publication.

pick-up truck scale (log$=3-4) geology papers generally have 1-5 authors.

Cruise-scale oceanography papers (log$=5-7) generally have up to a few dozen authours.

Spacecraft-scale (log$=8-9)planetary papers can have up half a page of authors.

Thus, big projects actually have fewer authors per dollar. Otherwise, space missions would have between 10,000 and 50,000 authors.

Dr. Lemming said...

Sorry, that should be 100,000-500,000 authors for space missions.

And 100-500 for cruises.