maps, territories, and discretion

Lorcan Dempsey mentions in passing that decisions of which manifestations belong to the same work set is”discretionary at the edges”:

A note on ‘discretionary’. We cluster stuff based on aggregate cataloger choices. I like Tim Spalding’s characterization of the ‘cocktail party test’ in a blog entry about works and LibraryThing.

Regarding ‘discretionary’, I think this is exactly right. It’s important to note that the ‘work set’ is a subjective and contextual choice, not some objective piece of data waiting to be discovered. But that doesn’t mean it’s useless, it’s very important because (in Western culture at least?), the concept of ‘work’ exists, and is of value to users — it’s socially constructed, its got grey edges, reasonable people may disagree in edge cases, but that doesn’t mean it doesn’t exist or isn’t useful!  (For instance, when a patron comes in asking for a copy of Hamlet, try telling her “Sorry, there’s no such thing as ‘Hamlet’, please come back when you can tell me a particular edition published in a particular place at a particular year that you want.”  Ha!)

The FRBR report says: “The concept of what constitutes a work and where the line of demarcation lies between one work and another may in fact be viewed differently from one culture to another.”  Quite right.

(This element of ‘discretion’ is present to some extent in ALL models of reality — and our bibliographic description is indeed a model. And always has been, even if it hasn’t always been formalized, even if it’s been based on traditional implicit shared understanding, not spelled out. The ‘map’ is never the ‘territory’, just a useful abstraction/approximation, with certain discretionary choices made as to be useful to a certain community/context).

So traditional cataloging tries to make these work distinctions (to the extent that they are implied in AACR2 choices like ‘uniform title’, and hopefully more explicit in RDA, but I can’t say for sure) by setting out precise instructions meant to result in choices that match that cultural determination of ‘work’.

LibraryThing tries to do it instead by just relying on members of that culture using their intution, and averaging out everyone’s choices and relying on them to reach consensus through discussion.

Neither is more ‘correct’, and neither is more ‘FRBR’, just two different approaches to trying to create a collective decision about work sets that is useful to users.  Both are discretionary and subjective.

Sometimes this is hard for those in the library community to understand; we seem to attract people who want there to be a ‘right’ and ‘wrong’ answer, who want “but is it REALLY the same work?” to be based on some kind of objective reality with an absolute discoverable ‘correct’ answer.  Sorry, that’s just not reality, modelling reality is inherently full of discretionary choices. But we can, as we traditionally do in library cataloging, set out rules and guidelines to make it as likely as possible that different people at different places will make the same choices.

But we really ought to explain why those rules and guidelines are what they are, and what the goal is. In order to better allow catalogers to use their professional judgement to make the choices most likely to accomplish that goal.

AACR2 rules that kind of sort of provide guidelines for making ‘work set’ decisions, but couch them in terms of simple orthographic decisions for ‘uniform title’, without ever even mentioning that this is really a decision about work sets — gee, it definitely doesn’t help us understand what we’re doing, and try to do it to serve the users as well as possible.  (It also doesn’t help that ‘uniform title’ is meant to express (at least) several other things in addition to ‘work set’ — we really need to record the ‘is in work set X’ decision as it’s own discrete reconstructable data element.) Anyone know if RDA improves on any of this? I still haven’t had the fortitude to try and make it through the RDA draft.

Work-centric or manifestation-centric display?

Lorcan also points out that:

Interestingly, Goodreads and LibraryThing seem to default to a work-based view: the entry is at the work level…   Amazon seems to default to a particular ‘manifestation’ or ‘expression’… Google Books seems to do something similar…. Worldcat.org is more like Amazon and Google. At the moment, it aims to show the most highly held member of a work set in a result, and then link to other editions from that…

I’d be interested if Worldcat is considering trying to make a ‘work’ view the default ‘landing page’ from a search, a bit more like LibraryThing. I suspect this would actually be of more general use than the library legacy practice of always showing individual manifestations as search ‘landing pages’.

Lorcan says: “There are reasons for taking these various approaches and each service make decisions based on what it is trying to do, and the view it takes of its user interests.”  Certainly true as far as it goes — but I’ve never seen a written out clear analysis of what the reasons for the traditional library manifestation-centered display are, what they are trying to accomplish, what user interests we believe they are meeting.

I suspect that in fact this choice isn’t based on any actual clearly thought attempt to meet certain user interests — but instead just because we’ve always done it that way. Because in the card catalog world it was impractical to do otherwise. And in the online world, it takes a bit more work to do otherwise.  Not because doing it this way actually is necessarily optimal for meeting identified user interests.

Please note that I’m not saying that we should ‘catalog at the work level’, whatever that would mean. Our cataloging practices certainly still need to describe manifestations, and there need to be different records for different manifestations. (On the other hand, something like subject cataloging probably is best done once at the work-level, not duplicated effort for every manifestation.) But a work-centric display can still be provided — if there is sufficient data recorded to allow software to reconstruct cataloger decisions about work-set groupings!  Current practice makes this difficult.

(And note that’s why Amazon and Google don’t have workset-centric displays. They don’t have the data to do it! Even Google’s vaunted algorithmic prowess can’t, apparently, determine work set groupings reliably enough to make a work-centric display. At least not within the resources Google is willing to throw at the problem. LibraryThing can do it because of volunteer human labor!  Library cataloging theoretically relies on such human labor, and we certainly spend an enormous amount of person hours in such labor — but don’t actually capture the fruit of that labor in unambiguous enough form to make it easy for the software to take advantage of. Shame on us.)