Wednesday, April 26, 2006

CHI Session -- Privacy 2 I'm in another CHI paper session on privacy issues. Hi session -- Privacy I'm in the CHI session on Privacy. The first paper is on incidental informational privacy. The scenarios is that you're at work, or in an airport gate area, and someone might look over your shoulder. They surveyed a broad set of users on their preferences and behavior. Not any real surprises here, just some good quantitative results that reinforce intuition. Second paper: Being Watched or being special. They reference a study from a study from the '70s that shows that people are much more willing to comply with impositions if a reason -- any reason, no matter how absurd -- is given. Their study shows that this extends to privacy. Fascinating.   12:32:44 PM    comment []

CHI Panel -- Tagging I'm in the CHI panel on "Why Taggigng systems work." This is really frustrating to watch. There are representatives from Yahoo/Flickr, Google, and various research institutions on this panel, all trying to define tagging, but REALLY trying to define tagging in a way where it's more important and signficant than just metadata. In essence, they're trying to define the "tagging phenomenon" while skirting around the fact that they all have a vested interest in tagging being an important phenomenon with long-term staying power. The one useful point raised is that in contrast to prior metadata efforts that were really designed around archiving and re-dscovery, tagging is largely focused on distribution (though certainly has an IR use too). Out of this has grown Luis von Ahn's work on cooperative community tagging (and how to use games to do this). I'm a big fan of Luis's work at CMU. Interesting observation/question from the audience: it seems like you need to be a "tag devotee" and pretty religiously do it to get a lot of value out of it. (panel answer: there's a fair amount of value just as a consumer for others' tags, e.g. Wikipedia) "man on the street" video, surveying people on their own filing/searching habits. What would get people to spend 30 minutes a day tagging web sites? Two most common answers: money (i.e. getting paid to do it) or "nothing." Interesting insight from George Furnas, University of Michigan: people overestimate their own ability to tag items accurately, and underestimate a group's ability to come up with a good diverse set that represents the object well. Furnas is definitely the star of this panel: he has a great historical perspective and a thoughtful approach that goes beyond the obvious memes of the tagging community (something the other panel members are having trouble with). Good audience question: will tagging make it outside of the community? Will out mothers ever tag things? (the moderator punted; he wants to come back to it at the end of the session) A panel member cited a UC Berkeley study that showed that tags are very similar to dialects: well-connected groups of people quickly converge on common sets of tags. When asked where tagging will go, really no clear ideas. Except fr one panel member who thinks we'll end up tagging (and thereby judging) people. An audience emember brought up that amazon added tagging to product pages a couple of months ago and it was a total disaster. A panelist said that it's Amazon's fault because the page is too busy. (another member jumped in and also blamed the UI) A third panelist is suggesting that the implementation was too eglaitarian -- not only could everyone tag, but everyone could define new tags. Question from a panelist: does tagging scale up to large, heterogeneous groups? the panelists seem to say "no" and I would suggest that this might be a more general indictment of social software systems: they almost never scale to large-scale, heterogeneous groups. recurring point that relates to this: one universal, flat terrain for tags probably doesn't work. You need to think about clusters of tags (potentially overlapping) and hierarchies. In other words, in classing Internet form, the tagging community has just rediscovered IR, taxonomies, and semantic hierarchies. Audience question: how many tags to people associate with an item? On delicious, the average is two (not surprisingly, that's the same as the number of words people type into a search box on MSN Search or Google).   11:15:24 AM    comment []

CHI Session: Mashups I'm in the CHI panel on Mashups. The BBC Backstage guy is giving a "Basics" talk on what Mashups are. Why do developers get involved in building mashups? new business opportunities it's cool they're frustrated with missing features/abilities in what the main provider supports to get noticed. BBC Backstage is BBC's developer network for supporting third parties creating mashups with BBC's data. They only support non-commercial use, and stress that all intellectual property remains solely with BBC. They offer broadcast schedule data, audio and video archives, plus travel data for the UK (train, road, etc.) BBC launched today reboot:bbc.co.uk, a competition to re-design the BBC home page. Cool idea. The Google guy is talking about the technical underpinnings of mashups. and why AJAX and lightweight feed protocols make it much easier to do mashup web apps. The data sources are growing faster than specific UI services are, which is a problem at one level and certainly exacerbates UI consistency issues since each mashup developer needs to roll their own. A good question from the audience about how to address accessibility issues for AJAX applications and machups in general. Not a lot of good answers to questions; mostly a lot of "good question, there are people thinking about that, no answer today." The discussant is talking about the privacy and security issues behind mashups. For example: do mashups make it really easy to develop a phishing site? Another issue: authentication for mashups. If you go to a mashup site and type in your password for another site, how do you know what's really going on behind the scenes? Will we see the return of Passport? or will Infocard pick up quickly, or will Liberty Alliance finally get going? Will SSL be required? (is that too costly in terms of getting an SSL certificate from Verisign?) The discussant is suggesting that mashup developers should develop more like enterprise developers. The Google guy just said that we need to be careful not to put too much burden on mashup developers to "do things the right way." and we should look for technical solutions instead. (my editorial view: there is a natural tension here, but if we really want mashups to take off, the responsibility needs to be both on the mashup enablers as well as the mashup developers) Is there a separation between mashups on Web sites vs. cell phones? The BBC guy says no. Some audience questions around the intersection of "citizen journalism" and mashups, and the issues of accuracy, authenticity and reliability of information. Also if there are errors, how do we build a feedback mechanism from end-users through mashups back to the original data source providers? Another audience question: mashups are a developer phenomenon today. Is there any chance to make it an end-user phenomenon? What would those tools look like? The Google guy thinks that it will happen eventually, but will just take time. My takeaway: the discussant (Hart Rossman, SAIC) has thought far more deeply about the issues behind mashups than either the BBC or Google guys. Mashups are very very young, and the hype has masked a number of severe limitations. We've seen a set of relatively simple mashups where the end-users cna remain anonymous (like layering data on top of maps) and that maps (no pun intended) well to 3 of the 4 reasons stated above why mashups are getitng built: coolness, frustration, and to get noticed. The real business opportunities, in order to be realized, will require actually tackling the hard issues, and we'll have to see if and how that happens -- or if not, how quickly mashups dies as just one more fad. I'm also disappointed at how little discussion there really was about the HCI issues related to mashups -- other than to point out that the HCI/usability community is not at all involved today.   7:37:15 AM    comment []

© Copyright 2006 Kevin Schofield. Last update: 5/1/2006; 8:59:58 PM.