The Semantics of Tagging - Social Semantics: The Search for Meaning on the Web - page 119

shown that tagging distributions of heavily tagged resources tend to stabilize into
power law distributions and present a method for detecting power law distributions
in tagging data, and seen the emergence of stable power law distributions as an
aspect of what may be seen as collective consensus around the categorization of
information driven by tagging behavior. Thus groups of tags are indeed an adequate
candidate for a notion of Fregean sense.
However, one could argue that the stabilization is just a mere artifact of tag
suggestions. Tag suggestions are when a tagging system, instead of letting the user
tag the resource, automatically (as the product of some algorithm) presents a list of
'suggested' tags for the user. The user can then easily accept these tags or choose
through them, rather than choose their own. This could lead to the stabilization of the
tagging system not as a product of the actual collaborative sense-making of users,
but as an artificial and predictable result of the tag suggestion algorithm. However,
the reasons behind the emergence of a power-law distribution in tagging systems
are yet unknown, although explanations fall into two general categories. The first of
these explanations is relatively simple: the tags stabilize into a power-law because
users are imitating each other via tag suggestions put forward by the tagging system
(Golder and Huberman 2006). The second and more recent explanation is that in
addition to imitation, users share through a similar tag generation procedure based
on the information on the webpage, most likely because the users have the same
background knowledge (Dellschaft and Staab 2008). However, drawing these two
influences apart has not yet been tested scientifically, which we will do. First let's
inspect the existing explanations for tagging stabilization more deeply.

5.3.1 Models of Collaborative Tag Behavior

5.3.1.1 A Simple Model: The Polya Urn

The most elementary model of how a user selects tags when annotating a resource
is simple imitation of other users. Note that 'imitation' in tagging systems means
that the tags are being reinforced via a 'tag suggestion' mechanism, and so the
terms imitation, reinforcement, feedback, and tag suggestion can be considered
to be synonymous in the context of tagging systems. The user can imitate other
users precisely because the tagging system tries to support the user in the tag
selection process by providing tag suggestions based on tags other people used
when tagging the same resource. There are minor variants of this theme, such as
the possibility of using a combination of tags of other users in combination with a
user's own previously used tags. In most tagging systems like del.icio.us these tag
suggestions are presented as a list of tags that the user can select in order to add
them to their tagging instance. The selections of tags from the tag recommendation
forms a positive feedback loop in which more frequent tags are being reinforced,
thus causing an increase in their popularity, which in turn causes them to be
reinforced further and exposed to ever greater numbers of users. This simple