Jesse James Garrett (from Adaptive Path) talked about user-generated information architecture ("Beyond Tagging"). The problem designers have today is that they can't always know the words people will use or the groups people will prefer. Card-sorting techniques is a primitive low-tech way to do the job. In the last few year, new approaches were used: instead of buidling architectures, build systems for users to build their own architecture (explicit activity): tags (user-defined keyword metadata), navigation devices (expose other people's keywords). BUT there are problems: insider language, "everybody use this tag" is a poor substitute for controlled vocabulary, is the most popular tag, by definition, best? Besides, the tag relies on people good well and sometimes the system is undermine with tag spam (tag things with irrelevant but popular tags), tag bombing (mistag content to make a statement about it.
So how can this could be improved? Amazon defies the conventional wisdom about information architecture (theh fact that users create mental maps of sites as they navigate (formulaitng their own classification system)). Amazon process an enormous amount of data: it knows everything you do with their website and they're able to generate architecture for you, to connect you affectively with their content.
The next step is to make this algorithm transparent to the users; to do that we need better data: about content that we are serving and about the users. The second one require another approach than usability testing ("usability testing are canes for the blind"): we need to treat EVERY visit into a usability test and every user in a test subject (to track effectiveness of their algorithm). And more sophisticated analysis of stats are needed to lead to more meaningful results than just most accessed directories. Unfortunately, the presenter did not go further explaining how this could be achieved, and the discussion stayed at the information architecture level (I know that's the topic of the talk but I would be interested into deeper things such as practices investigation or queries analysis over time).
Then, Bruno Guissani started by saying that he is tired that old media will be killed by new media or that journalists are not reliable. To him, the discussion should be more about where both will converge. This is something I fully agree with (old media does not disappear but have a peculiar relationship with new media such as blogs).
He presented the example of the swiss magazine "L'Hebdo" which, during the french "banlieue" riots in 2005, instead of sending journalists there and quickly getting back home; they sent almost every journalist on a weekly rotation (sth like 10 days per person) and they reported news/interview/insights on a blog. They rented a small studio on the groundfloor in the "banlieue". This impacted the way journalists worked (reconnecting journalism and field investigation) and also how others newspaper looked at this experience (and in the end tell their stories ina better way). But it also impacted the city itself, and now people from the city voluntereed to take care of the blog: it's the first media in this city now (there were no newspaper there previously). The whole thing is financed through a book that compiled the stories coming from the blog.