“Friendfeeds for Science” pt II – Design ideas for a research focussed aggregator

This post, while only 48 hours old is somewhat outdated by these two Friendfeed discussions. This was written independently of those discussions so it seemed worth putting out in its original form rather than spending too much time rewriting.

I wrote recently about Sciencefeed, a Friendfeed like system aimed at scientists and was fairly critical. I also promised to write about what I thought a “Friendfeed for Researchers” should look like. To look at this we need to think about what Friendfeed, and other services including Twitter, Facebook, and Posterous are used for and what else they could do.

Friendfeed is an aggregator that enables, as I have written before, an “object-centric” means of interacting around those objects. As Alan Cann has pointed out this is not the only thing it does, also enabling the person-centric interactions that I see as more typical of Facebook and Twitter. Enabling both is important, as is the realization that all of these systems need to interoperate effectively with each other, something which is still evolving. But core to the development of something that works for researchers is that standard research objects and particularly papers, need to be first class objects. Author lists, one click to full text, one click to bookmark to my library.

Functionality 1: Treat research objects as first class citizens with special attention, start with journal papers and support for Citeulike/Zotero/Mendeley etc.

On top of this Friendfeed is a community, or rather several interlinked communities that have their own traditions, standards, and expectations, that are supported to a greater or lesser extent by the functionality of rooms, search, hiding, and administration found within Friendfeed. Any new service needs to understand and support these expectations.

Friendfeed also doesn’t so some things. It is not terribly effective as a bookmark tool, nor very good as tool for identifying and mining for objects or information that is more than a few days old although paradoxically it has served quite well as a means of archiving tweets and exposing them to search engines. The idea of a tool that surfaces objects to Google is an interesting one, and one we could take advantage of.  Granularity of sharing is also limited, what if I want slidesets to be public but tweets to be a private feed? Or to collect different feeds under different headings for different communities, public, domain-specific, and only for the interested specialist?

Finally Friendfeed doesn’t have a very sophisticated karma system.  While likes and comments will keep bringing specific objects (and by extension the people who have brought them in) into your attention stream there is none of the filtering power enabled by tools like StackOverflow. Whether or not such a thing is something we would want is an interesting question but it has the potential to enable much more sophisticated filtering and curation of content. StackOverflow itself has an interesting limitation as well; there is only one rank order of answers, I can’t choose to privelege the upmods of one specific curator rather than another. I certainly can’t choose to order my stream based on a persons upmods but not their downmods.

A user on Friendfeed plays three distinct roles, content author, content curator, and content consumer. Different people will emphasise different roles, from the pure broadcaster, to the pure reader who doesn’t ever interact. The real added value comes from the curation role and in particular enabling granular filtering based on your choice of curators. Curation comes in the form of choosing to push content to Friendfeed from outside servces, from “likes”, and from commenting. Commenting is both curation and authoring, providing context as well as providing new information or opinion. But supporting and validating this activity will be important. Whatever choice is made around “liking” or StackOverflow style up and down-modding needs to apply to comments as well as objects.

Functionality addition 2: Enable rating of comments and by extension, the people making them

If reputation gathering is to be useful in driving filtering functionality as I have suggested we will need good ways of separating content authoring from curation. One thing that really annoys me is seeing an interesting title and a friendly avatar on Friendfeed and clicking through to find something written by someone else. Not because I don’t want to read something written by someone else, but because my decision to click through was based on assumptions about who the author was.  We need to support a strong culture of citation and attribution in research. A Friendfeed for research will need to clearly mark the distinction between who has brought an object into the service, who has curated it, and who authored it. Both should be valued but the roles should be measured separately.

Functionality addition 3: Clearly designate authors and curators of objects brought into the stream. Possibly enable these activities to be rated separately?

If we recognize a role of author, outside that of the user’s curation activity we can also enable the rating of people and objects that don’t belong to users. This would allow researchers who are not users to build up reputation within the system. This has the potential to solve the “ghost town” phenomonen that plagues most science social networking sites. A new user could be able to claim the author role for objects that were originally brought  in by someone else. This would immediately connect them with other people who have commented on their work, and provide them with a reputation that can be further built upon through taking on curation activities.

This is a sensitive area, holding information on people without their knowledge, but it is something done already across indexing services, aggregation services, and chat rooms. The use of karma in this context would need to be very carefully thought out., and whether it would be made available either within or outside the system would be an important question to tackle.

Functionality addition 4: Collect reputation and comment information for authors who are not users to enable them to rapidly connect with relevant content if they choose to join.

Finally there is the question of interacting with this content and filtering it through the rating systems that have been created. The UI issues for this are formidable but there is a need to enable different views. A streaming view, and more static views of content a user has collected over long periods, as well as search. There is probably enough for another whole post in those issues.

Summary: Overall for me the key to building a service that takes inspiration from Friendfeed but delivers more functionality for researchers, while not alienating a wider potential user base is to build a tool that enables and supports curation rating and granular filtering of content. Authorship is key, as is quantitative measures of value and personal relevance that will enable users to build their own view of the content they are interested in, to collect it for themselves and to continue to curate it for themselves, either on their own or in collaboraton with others.