tagging, communities, vocabulary, evolution

Really interesting discussion of community tagging and how best to facilitate it

This paper explores some factors that influence user tagging behavior, using the MovieLens site. Unlike most other papers on tagging, this one actually compares data across different versions of the same system. Here are the versions:

1) Users only see their own tags. 2) Users can see a random set of community tags. 3) Users can see the most popular community tags. 4) Users see recommended tags applied by the community to the item and similar items.

Observation (from paper): number of tags given by a user approximates a power law distribution, with a few users generating most of the tags.

Three types of tags: factual, subjective ("stupid"), personal ("toread"). (Not sure how many subjective tags Flickr has.)

MovieLens is more like del.icio.us than Flickr, since users don't "own" the movies and tags are bags instead of sets.

Habit and investment (tags given in the past) influence user tagging, and this influence grows stronger over time. Community influence also affects user tagging.

Showing users popular tags encourages them to give factual tags. (I wonder if this means the ESP game is overly careful.)

Actual tags also get reused more by users who see popular or recommended tags.

It seems that most users did not consider tags useful at all (though this is for movies and may not apply to other domains), though they considered factual tags more useful than the other two types.

This is a case study on tagging in the movielens system (very interesting). One aspect to investigate is the tag selection algorithm.

Awesome article -- rigorous and directly relevant for information architects and designers.

Social proof states that people act in ways they observe others acting because they come to believe it is the correct way for people to act [5]. For example, Asch found that people conform to others’ behavior even against the evidence of their own senses [1].