« When outsourcing makes sense | Main | From Salford Quays to synthetic worlds »

November 16, 2006

The "social" in social tagging

One of the sessions I regret missing (prior commitments beckoned) at DC-2006 was a "special session" on "Social Networks: Tagging and Dublin Core metadata" facilitated by Liddy Nevile. From what I gathered from chatting with a few people who did attend, looking at some of the presentation materials posted and observing some of the email exchanges afterwards, it seems to have been one of the more successful meetings to have taken place during the conference. As a result of the meeting, a proposal was made for the formation of a "community" - a forum for open discussion - within DCMI to explore the relationships between the "traditional" approaches to metadata creation and use and the more informal, community-oriented approaches to metadata that have emerged, particularly through the use of "tagging" in the context of various social software systems. (Details of the DCMI Social Tagging Community are now available.)

One of the first points of discussion was the name of this community and in particular whether that name should refer explicitly to "social tagging", rather than simply "tagging". This post is a slightly extended version of a message I sent to the DCMI Social Tagging mailing list on that subject.

Liddy posted a short piece to stimulate discussion, and drew a distinction between the creation of metadata by "trained cataloguers" (using the example of MARC records created by librarians) and the creation of metadata by "ordinary people", non-experts without training in cataloguing practices (using the example of Dublin Core), and suggesting a similarity between the latter and "tagging".

I think this does highlight one facet of "tagging" which is important: the simplicity of assigning tags. To tag a resource, I choose whatever tags I wish to associate with the selected resource. I don't have to worry about whether I'm using the "right" tag, and I don't have to scrutinise DCMI's documentation to unravel the mystery of when I should use "Bristol" as a value string with the dc:subject property and when I should use it as a value string with the dc:coverage property (well, unless I adopt some hare-brained convention for constructing tags which itself relies on that distinction!)

(As an aside, while I'd agree that in some cases DC metadata has been created by people who are not trained cataloguers, and creating DC metadata is generally simpler/less complex than creating MARC metadata, I'd qualify that by saying that this begs the question of what we mean by "DC metadata". If we mean metadata created using the description model described by the DCMI Abstract Model, then the degree of complexity involved in the creation of instances varies depending on the particular DC application profile being used. A DC application profile may be relatively complex and effective metadata creation may well require some degree of familiarisation/training.)

However, I'm not sure that simplicity is the distinction which people are seeking to capture through the inclusion of the "social" adjective. I think the intent is not to indicate the simplicity of the "tagging" operation, or the level of expertise required, but rather to emphasise that the operation takes place in the context of a communal or collaborative system. The levels of training or expertise of the taggers within the community, and whether there are variations in those levels or not, are a characteristic of the community, rather than of tagging itself. And while most of what we now refer to as tagging does take place within such communal systems, I don't think that is necessarily the case: tagging is often social but it may not be.

In theory, I could engage in "tagging" within a system in which I was the only user. I could run a del.icio.us clone on my laptop, accessible only to me on my user account on that machine, and I could post entries and "tag" resources within that system.  In this scenario, I'm certainly performing the "tagging" operation. But there is no communal or collaborative aspect to that operation. I'm not sharing my collection of entries (including my tags) with anyone else, and I'm not looking at entries (including tags) from the collections of other individuals. No-one else is analysing or using my tags and modifying their tagging behaviour based on that experience, and I'm not analysing or using anyone else's tags, and modifying my tagging behaviour based on that experience. Retrieval within the system is based only on my own tag set. The "description" of a resource constructed in this way is created through my input alone. This is "tagging", certainly, and it may be very useful to me for my personal information management, but it's not "social tagging".

If I was doing a similar thing in a system that was hosted on our organisational intranet, and a few colleagues were also posting entries to their own collections and applying their own tags to resources, and we were browsing each other's collections and using each others' tags, then a communal element is introduced. Even if there are only a handful of contributors, there is now a social dimension to the operation. And typically social tagging software makes this explicit both at the point of tagging (by, for example, offering the tagger suggestions for tags based on the tags used by other contributors) and at the point of retrieval/use (browsing the community's tag set as well as my own tag set). We are users of the tags we ourselves assign, and of the tags created by others within the community. And community tagging behaviour conditions our experience of retrieval and may condition the tags we choose.  And perhaps more interestingly, these multiple taggings by different contributors mean that there is now a "description" of a resource created through the aggregated input of multiple contributors. This is now a form of what I would call "social tagging". In this particular scenario, the "community" is small and probably quite homogeneous in terms of their aims, experience/interests, use of terminology, and so on (and probably more broadly of culture and language). Even within such a small group, different members might "engage" with that "community" to a greater or lesser degree - we might choose to try to converge on a shared set of tags across the community (e.g. I might tag something as "semweb", discover later that my colleagues use a tag "semanticWeb" and then edit/replace my tag accordingly) or we might choose to ignore each others' tagging conventions! - but there is a social dimension both to the tagging operation and to the subsequent use of those tags in retrieval.

And of course that scenario extends to the more familiar open, global, Web-based systems (like del.icio.us and so on) where the community of participants is large and heterogeneous, with wide variations in aims, experiences, culture, language etc. They are all potential users of my tags and I'm a potential user of theirs. In this community, yes, both experts and non-experts are contributors.  And the process results in "descriptions" which are the result of combining the inputs of many - in some cases hundreds or thousands of - different contributors. These descriptions are social, communal constructs, rather than the creation of any one individual participant.

In these scenarios I'm not making any assumptions about the "ordinariness" (or otherwise! ;-) ) of the participants or about their levels of expertise or training. A community of "trained cataloguers" might engage in social tagging and "ordinary people" might "tag" in ways which are not "social".

Turning briefly to Dublin Core metadata, I'm not convinced that DC metadata creation has typically taken place within this sort of communal/collaborative context. And in that sense I'd probably say that DC metadata creation has typically not been a "social" process, and the "descriptions" created are not "social" constructs, or at least not in the way that those generated by tagging within a system like del.icio.us are. And we (the DCMI community) can probably learn from examining those explicitly social aspects of those processes and systems.


TrackBack URL for this entry:

Listed below are links to weblogs that reference The "social" in social tagging:


Given the attributes you've assigned to social tagging (collaborative development & assignment of descriptive terms, multiple users assigning terms to the same resource, public availability of terms for search), is there any reason not to consider WorldCat a social tagging application for the library community?

And if not, would the ways the library community handles collaborative development of tags suggest anything on how to improve consistency/quality of tagging outside the library community?

I've always seen the word 'social' in social tagging as a bit of a strange monicker. In the one sense, 'community' might be more appropriate, or 'shared', or 'negociated'.

I'd further note that a major advantage of social tagging is that it accepts the situated nature of the word 'expert'. In other words, whilst, as you point out, negociation of shared terminology is inherently a social/collaborative process (hey, it's an interaction!), the more interesting element of social tagging from my perspective is the 'shared representation' aspect of the beast -- shared representation as in socially shared, in the Vygotskyan sense.

There's an important way in which DC creation is ALWAYS a social process of one kind---the only point to it is if you are planning on sharing your metadata. Right? Or at least most of the point.

If you have your hypothetical delicious clone on your laptop, just for you, and nobody else can see it---there's a lot LESS reason to even consider using DC. What problem does it solve? The problems DC solves are usually about sharing, which is social. No?

The comments to this entry are closed.



eFoundations is powered by TypePad