Hana Shepherd and Amir Goldberg – Tagging politics and consensus
What are the possible consequences for a lack of consensus in tagging systems, ask Hana Shepherd and Amir Goldberg. Might there be bias in retrieval of sites that might systematically disadvantage particular types of resources or particular types of users from straightforward and intuitive online participation? Put differently, is folksonomy merely the tyranny of the majority in a new guise?
As you might have noticed in your last browse through the online world, more and more sites allow users to “tag” resources. From the more popular ones like the social bookmarking site del.ici.ous and the ever-present YouTube and Amazon, to the less-well-traveled information and data-organization sites like Furl, LibraryThing, and Citeulike, tags are everywhere. Tags allow users to attach their own keywords to online resources (websites, blogs, pictures, videos, books, etc.), theoretically allowing for more efficient and user-driven retrieval of resources. But sites that use tagging, or at least web developers interested in the much-hyped Semantic Web, have their eyes set on a higher goal as well. Tags on a site are examined in the aggregate, with the idea that they can be used to develop “folksonomies,” or user-driven metadata, as a way of characterizing data online by looking at the distribution of tag usage frequency. The possibility of efficient and user-generated metadata, as opposed to categories of concepts created by experts, generates much excitement, as tagging systems are supposed to allow much greater malleability and adaptability in organizing information than do formal classification systems, and thus present the possibility of better online navigation. As products of cumulative user effort, they harbor the promise of genuine spontaneous collective intelligence.
While this all seems quite promising in terms of the ability of any user to play a role in the development of how the web operates, a quick look at the assumptions behind such proposals reveals that all is not quite so rosy. Proponents of tagging systems argue that “groups of users do not have to agree on a hierarchy of tags or detailed taxonomy, they only need to agree, in a general sense, on the ‘meaning’ of a tag enough to label similar material with terms for there to be cooperation and shared value”. This “agreement on meaning” may be possible in some cases of tagging systems, particularly those where the user-community is quite homogeneous, or drawn from a similar subgroup, but may be quite problematic in other cases. A recent search of the tags on Amazon revealed that the tag “defectivebydesign” was one of the most popular, used for 1086 items by 520 users, but was meant to signal items that those users did not like for a variety of arguably idiosyncratic reasons. Presumably, there are many users who feel favorably (or even neutrally) about those 1086 items, but their tags might have been overshadowed by the enthusiasm of those who dislike them.
Meaningful tagging seems to necessitate consensus about the meaning of resources, signified by the tags used to describe them. But how can consensus be reached? In particular, we would like to highlight three issues for which there are no definitive answers: 1. How might we know whether there is consensus between users of a tagging system about “meanings” (i.e. pattern of relationships between tags and resources)? 2. Who becomes influential in forming consensus? 3. What are possible consequences for a lack of consensus in these online systems? Answering these questions seems to depend on how one thinks about the relative importance of preexisting social groups and boundaries in shaping the “meaning” that individuals attach to resources. There has been documentation of inequality in both access to and type of use of the web based on preexisting social groups and divisions (gender, race, age, etc.) which may, if those social divisions are important to how users tag, generate several possible “meaning systems” within any given tag distribution for a resource. Yet any individual or social group predispositions are constantly renegotiated through interaction with other users, whether by discussion and information sharing commonly facilitated by tagging websites, or through features of the sites that encourage the generation of shared “meaning” of resources, such as the salience of previously used tags.
What would consensus in online tagging sites, or lack thereof, look like? Some recent work has examined the distributions of tags within resources for deli.cio.us and suggested that the distribution seems to stabilize, or reach a distribution that is fairly impervious to future tagging, after a substantial number of users tagged a website. But is the stabilization of a distribution enough to say that consensus has been reached? If there were pre-existing subgroups of users who were attaching different sets of meanings to the resources, but whose numbers were small, would they be able to be detected? Would deriving content from distributions of tags mean looking only at the most popular tags to the exclusion of those less frequently used? Do subgroups who may be developing their own independent tagging dynamics become noise which is suppressed in the process of focusing on the heavily-used tags?
Moreover, what are the possible consequences for a lack of consensus in tagging systems? Might inattention to subgroups with their different meaning systems for resources cut off entire groups of users from facile participation in web searches if a “folksonomy” based on what is perceived to be “shared meaning” or consensus among users is developed? Might there be bias in retrieval of sites that might systematically disadvantage particular types of resources or particular types of users from straightforward and intuitive online participation? Put differently, is folksonomy merely the tyranny of the majority in a new guise?
Tagging systems, we suspect, do not only reflect on the idiosyncrasies of online communities. Many of the mechanisms involved – how groups bring and negotiate meaning, how meanings become dominant, what the consequences are for those who “lose out” – seem to resonate with concerns of social inclusion and exclusion which are as old as the study of society itself. If tagging systems continue to proliferate, it seems that sorting out these questions about consensus will be extremely important in ensuring the democratic promise of participation in the creation of online content, and perhaps offer new ways to understanding how meaning and patterns of social relations interact in reproducing social inequality.
0 σχόλια »