At the JEDI 2 workshop held in
mid 2008* we spent quite a lot of time envisaging how cosmology might be different in 2020. One aspect of that was the way in which we would do research in 2020 and I would like to discuss one of my proposals at JEDI 2 here, which is in some vague way a follow-on from my earlier post on video abstracts for papers.
To generalise shamelessly, I find that most papers have rather poor introductory value (including my own no doubt), in the sense that new students will not be much the wiser for having read them. In fact one could be harsh and say that most introductions are written for nobody – experts don’t need or read them and new students can’t understand them. Why? Well, the introduction is often something we write largely because journals demand them, not because we as authors particularly care about doing a great job. So what one has instead are hundreds of poorly written, ineffective introductions to the same topic. Students move from one paper to another trying to gleam enough information to piece together a viable overview of the field.
What can we do? Here is a concept proposal – Library-based, object oriented paper writing – which I don’t even fully endorse yet myself. But I think it is interesting and perhaps even inevitable, so worth discussing informally here.
Let me start by making an analogy. In scientific computer programing it is common to make use of libraries. Classic examples are LAPAC and Numerical Recipes. Why do people use these libraries instead of writing their own code? There are many reasons, but key ones include (1) they are tested and correct, (2) they can be “easily” integrated into your own code and (3) they allow one to get answers without having to have detailed knowledge that particular algorithm (they are “black/gray” boxes). While using libraries may not be a good idea in some cases, they save time both in coding and running and, if used properly, give better results.
The question is, could we do the same for paper writing?
My proposal is that instead of only being able to cite whole papers, perhaps we should move to a model of citing objects. An object might be a section of text (e.g. a great introduction to dark energy), an image, a video, a piece of code. In this framework, writing a paper would more naturally be seen as linking objects, some of which were produced by someone else. Using an object would attract a citation and hence would give credit to the author. In turn, one’s own paper could be sub-cited.
Writing a paper would involve choosing a suitable introduction from some online library and dragging and dropping it into your paper with appropriate citation. One could even have multi-level branched text. For example on might have a sentence as follows:
Dark energy is one of the most pressing problems in modern cosmology (basic introduction: +, advanced introduction +)
Where “+” indicates the possibility of expanding the object one has imported. Of course this immediately means hard-copy printing is more tricky, but how many people actually read printed papers these days?
Underlying this idea is the notion that I no longer consider my whole paper to be mine. Rather the sea of published work becomes available to use as appropriate to make my paper as good as possible, with credit going where it should.
In the industrial world this is nothing new of course. Manufacturers use sub-components bought from other companies all the time. The difference is that instead of paying for the component, we would simply cite them. We have a new, more subtle academic economy.
An interesting by-product of this would be specialisation (which may be a good or bad thing). People could get large numbers of citations by writing wonderful introductions or descriptions of subjects. Students would get introductions that actually explained things to them properly (and which contained excellent sets of references, meticulously maintained and updated) and researchers would be freed from the torture of writing how important dark energy is blah blah.
Obviously this has lots of implications, too many to discuss here, but let me close with some final thoughts:
(1) Finding the appropriate part of a paper would be much easier. Instead of simply being told that the result is somewhere in that article of 25 pages, we would know it is in section 3.2.
(2) How could this be implemented in practise? I think the arxiv could be expanded naturally to allow sub-citing of articles. A new referencing language would probably be required, but latex trivially allows referencing of sections and equations in one’s own document. It would need to be extended to allow for referencing in other papers, a nontrivial task of course.
(3) Will this actually work or is paper writing so bespoke and unique that it would save no time or effort and not lead to improved learning for students? I am not sure, but I think it would…
I would be interested to hear what people think…
Update 19/5/2011: Quite by chance I came across this channel on Youtube by the Journal of Number Theory where authors are encouraged to post video abstracts of their papers. Watching a few I couldn’t help thinking that somehow they have missed the point a bit, but equally, perhaps, I may be too fond of superficial entertainment and a little too averse at looking at academics when I already have to look at them everyday!
* Date corrected 19/5/11.
Update: 12 June 2011: An interesting related post is available here on nano-publications and the future (or lack thereof) of journals. The inspiration is here. BB
Update 8 February 2012: One aspect of this whole issue is now addressed by the Astrophysics Source Code Library which now allows people to cite code as standalone objects on their own without needing a reference paper. Excellent. Seems appropriate somehow that the object-oriented model of research should apply first to code! BB