Amy Bruckman at George Institute of Tech says that researchers knock the use of the word “community.” We ought to use a prototype model for defining communities, using Eleanor Rosch’s idea of prototypes, she says. [Go Rosch!]. Our prototypes for communities vary by “genre,” e.g., Flickr is one type of community and so is the WELL; we understand them in relation to different prototypes. The work she proposes going forward would try to discern the relevant differences among the prototypes.

Robert Kraut at Carnegie Mellon’s CommunityLab Project begins by upbraiding the conference for using a panel of kids to discover what kids think instead of using the serious research that’s been done on the topic. He suggests using group theory to design online groups. For example, in many online communities, a small percentage of people (80:20, roughly) contribute. The “collective effort model” explans under contribution and the conditions that mitigate it. It says that “effort = outcome probability x outcome value.” It predicts you can increase contribution by make salient the uniqueness of contributions, increasing how much people like the group by making it more homogeneous, and make the benefits of contributing more salient. (This is based on well-established research, he says.)

He describes two experiments.

First: To get more people to contribute movie ratings, the system reminded posters of their uniqueness (identify a movie the person has rated but no one else has) and made the group more attractive by putting together people with similar tastes. Results: Uniqueness increased the number of discussion posts and movie ratings, but similarity in tastes depressed the posts. (The first finding supports the hypothesis, the second disconfirms it.)

Second: They sent email inviting people to contribute ratings. Some were told thjey’ve been invited because they have unusual tastes, others because they have typical tastes. Some were told that the more ratings they contribute, the better the system works for you, or for other people. Again, emphasizing uniqueness helped, but being told the number of benefits they would get depressed the results.

He concludes by looking at how useful it is to approach the design of online communities by starting with theory. He found that it inspired design features not often used and allows for reuse of principles.

Randy Farmer, now at Yahoo! but an early pioneer in Net communities, says that at last we’re ready to scale. But, he points out, there are problems with scaling communities, including spam, fraud and “intellectual property” issues.

Massively Multiplayer games have blazed one trail, Randy says, raising hard questions of virtual economies and property law: Are virtual goods property? He says Sony is enabling an actual cash market for Everquest.

Social networking has blazed another path. Because too much info was published with too little to do with it, selective disclosure is arising.

Challenges: Boundaries of identity and disclosure. How do we scale trust? Answer: “Hand the trust question over to the users.” He says Yahoo’s “This is spam” is an example of handing trust to users. And how will public-scale tagging avoid tagspam. How do we do “pagerank” for tags?

1. Tag rating: Positive reputation: “537 people tag this as penguin” Negative: People respond to a tag saying “This is not a penguin.”

2. Tagger reputation: As people are known for putting better tags on things, their tags count for more.

David MacDonald at U of Washington looks at “visual blogging communities” — sites with photos and a little text. (Example 1 2 3) His project has been archiving 9-25 sites of the 53+ he knows about. They picked 19 sites with at least 3 months of data. You can do content analysis (what’s in the picture?), ethnographic analysis (what is this picture about?), and interaction analysis (what’s the relation among a set of pictures).

He talks about preliminary results from the interaction analysis. He suggests some categories: Positional polay (glance, point), inmage stealing (re-use), theme (impromptu or planned group action), text in picture as a title (which forms the majority of the interaction). Then he gives great examples of how people engage in various forms of “conversation” in these photo sites via their images. (Molly Wright Steenson in the backchannel points to the Internet employee of the month page at Flickr.)

Fernanda B. Viegas at MIT Media Lab talks about a project visualizing the evolution of wiki pages. She’s been visualizing archives: email, usenet, chat. She shows work, with Martin Wattenberg, in visualizing wikis as a way of understanding the dynamics of the community that builds them. You can play with it yourself: HistoryFlow. She points to the abortion page, showing how quickly it gets restored. And she points to the chocolate page where a zigzag pattern indicates an editing war, in this case over whether there is such a thing as a “chocolate collage.” The patterns also show that the text of the people who start a page tends to be long-lived.

She says IBM is releasing the software and will likely also release the Wikipedia plugin.

[Excellent morning.]

[Technorati tags: scs2005 community]

Categories: Uncategorized dw