The Role of Tag Suggestions in Folksonomies

TL;DR

This study investigates how tag suggestions influence the distribution of tags in folksonomies, revealing that power law distributions emerge independently of suggestions and are subtly affected when suggestions are used.

Contribution

The paper provides empirical evidence that tag suggestions do not cause power law distributions in tagging systems, challenging existing theories based on imitation behavior.

Findings

Power law distributions form regardless of tag suggestions.

Tag suggestions cause a 'compression' of the power law distribution.

Power laws are not primarily driven by user imitation of suggestions.

Abstract

Most tagging systems support the user in the tag selection process by providing tag suggestions, or recommendations, based on a popularity measurement of tags other users provided when tagging the same resource. In this paper we investigate the influence of tag suggestions on the emergence of power law distributions as a result of collaborative tag behavior. Although previous research has already shown that power laws emerge in tagging systems, the cause of why power law distributions emerge is not understood empirically. The majority of theories and mathematical models of tagging found in the literature assume that the emergence of power laws in tagging systems is mainly driven by the imitation behavior of users when observing tag suggestions provided by the user interface of the tagging system. This imitation behavior leads to a feedback loop in which some tags are reinforced and get more popular which is also known as the `rich get richer' or a preferential attachment model. We present experimental results that show that the power law distribution forms regardless of whether or not tag suggestions are presented to the users. Furthermore, we show that the real effect of tag suggestions is rather subtle; the resulting power law distribution is `compressed' if tag suggestions are given to the user, resulting in a shorter long tail and a `compressed' top of the power law distribution. The consequences of this experiment show that tag suggestions by themselves do not account for the formation of power law distributions in tagging systems.