At the Library of the University of Amsterdam, we've done a small trial to investigate the merits of using a 'social referencing' service. In the trial, scientists from a specific research group used such a service for several months. With a panel discussion before and after, and logs during, we hoped to get some measurable results on the impact on their work of this. Not all went as we hoped, but the resulting report (url below) offers some useful insight in what this group of scientists found useful.
I'll highlight two points that may be of wider interest. First of all, our test group felt strongly about privacy. They liked to not just add cites to a system, and tag, rate or even comment them heavily for themselves. For sharing however, it was felt essential to have control over who could see what. Rating an article negatively could be helpful for direct peers, but who knows whether the author might decide over a grant in future? The participants wanted a clear, fine-grained control. Sometimes it would be fine to share the citation, but to limit the rating or tags to a certain group; some comments are meant private, some are for the whole world; and so on. Because of this, they voted to use BibSonomy at the start of the trial.
However, it turns out that ease of use is even more important than features, and the group considered the system not easy enough. To save the project, we switched to Citeulike, and extended the use period.
I'm curious what others think of this demand for detailed control over privacy-settings. In researching next generation research collaboratories, again I found scientists consider it paramount to be able to set the privacy level for each item themselves. Not through a helpdesk, ticketqueue and a sysadmin - themselves. The two main contenders to build such systems on top of, Sakai and Sharepoint, provide such a detailed rights-structure out of the box. So it's not pie-in-the-sky thinking: it is already out there.
The project page
Download report (pdf)
BTW: yes, we're working on a proper publication.