Guillaume Cabanac is conducting an ambitious experiment on the automatic evaluation of discussion threads on web annotation systems. Web annotation systems are an interesting premise that could make the web more interactive. The problem is that the results are often messy and unusable. This experiment is part of an effort to reduce this mess.
The ultimate aim is:
to determine if our algorithm (i.e. a computer program) is able to compute a “social validation” value that is close to human perception. As an application, this would enable readers to focus on annotations that have been confirmed by their replies, for instance.
The experiment itself asks the subjects to evaluate discussion threads as to whether they support proceeding arguments, as well as whether they themselves are confirmed of refuted. It will be interesting to see the results, as some of the threads are pretty complicated for a human actor. Of course there is the social aspect of the problem which is not addressed yet: that there will be more arguments from extremists than anyone else so their views will tend to be confirmed/supported.
So, if you are interested and have some time to spare, see for yourself.
Leave a Reply