Knowledge Organization! A mention of the subject leaves the average person scratching their heads. What exactly is it? In truth, attending a conference on knowledge organization left me little the wiser as to its exact place in the world. Knowledge Organization (KO) has a kind of double existence. It can be the low-level domain of cataloguers as they painstakingly and lovingly annotate and create individual library metadata records, or study the effect of blank space in a library classification catalogue; or it can start from the high-level attempt to manage all information seeking of whatever kind. Perhaps it was not surprising the presentations at this conference varied enormously in scope (and utility).
This conference was a bit of a contrast from the 750 attending the UKSG Conference in Glasgow; there were around 70 attendees, of whom only around half were there in person – but that did mean it was possible to chat to everyone, and that is, after all, the main appeal of conferences. Conferences at this scale have an atmosphere like no other.
What was included? There were 24 presentations, with no break-outs (that is, every participant sat through every talk: there was no escape for attendees). The presentations were a mixture of talks by established academics, consultants involved in commercial (or non-commercial) projects, and research students. Impossible to summarise all 24 presentations, so I’m restricting myself to broad themes.
The tendency of Knowledge Organization to explode into the entirety of all human, political and social history was especially apparent at some of the presentations by PhD students, who for some reason were zealously guarded by their supervisors in their talks, as if they might be too fragile for the cut and thrust of questioning by the outside world – and an ISKO Conference is hardly the outside world. To be honest, it wasn’t that kind of conference; everyone was very polite and friendly. The PhD presentations were impressive yet at the same time alarming: impressive because they attempted to identify something valuable, such as how different academic disciplines have different mental models, but at the same time alarming, because trying to capture in what way researchers think in chemistry, history, medicine and law is such a vast topic that even a PhD risks being superficial. Perhaps that is why the supervisors were so keen to protect the young researchers from questions. Other topics seemed incredibly narrow, such as studying social media comments for the National Library of Scotland (hardly the place where a lot of comments will be posted, I would have thought). Other PhD presentations seemed to reveal the challenge at the heart of any study of knowledge organization, looking at the entirety of human history only to reveal something rather obvious, for example, that what constitutes Jewish dietary law has varied widely over the centuries, not because Jewish texts change, but because interpretations of them vary dramatically over time.
But KO is changing. Most importantly, generative AI hangs like a heavy shadow over the entire subject of knowledge organization: do we need it at all? Can’t a machine do everything for us? Bringing Chat-GPT down to earth means ignoring the wilder claims for AI to change our lives instantly, but examining the much more narrow set of tools for corpus-based AI and making predictions based on largely statistical analysis of that corpus. You might imagine this is part of knowledge organization, but it doesn’t appear to figure highly on the agenda of many of the speakers.
You noticed the absence with some presentations that seemed to turn their face from AI, as, for example, in the keynote talk about “information behaviours during times of personal change”, and Deborah Lee’s talk about “exploring visualization within classification schemes”. Worthy though these topics might be, they left me questioning their significance compared with how AI threatens to transform the entire world of information seeking and retrieval.
Noticeable was a divide in the room (both of attendees and speakers) between those who were familiar with and who used AI tools, and those who did not. I was surprised that more than one of the academics made it clear they had little knowledge of AI (and it didn’t sound as though they were at the conference to gain this knowledge).
Incidentally, Lee’s talk was one of two presentations at the conference about the visual representations of knowledge structures, neither of which mentioned the term “infographic”, even if the term was raised in the questions after the talk. This, to me, summed up another potential problem of knowledge organization: it runs the danger of not noticing significant work in other fields, which happens not to be labelled “knowledge organization”.
Other presentations focused so much on the minutiae of knowledge organization that they overlooked the way that formal KO has been to some extent bypassed in real-life discovery systems. So a talk on “browsing ethnographic videos by freely faceted classmarks” (by no fewer than six authors) was, I think, of relevance only to a tiny minority. If the goal was to get the ethnographic videos noticed, there are better ways to do this than classifying them; try a few tweets and rely on serendipity, perhaps.
For me, the successful presentations were those that looked to link AI with knowledge organization. Paul Groth gave a wide-ranging talk about trying to combine “traditional” knowledge organization techniques with LLMs (large language models), and impressed by his optimism and willingness to combine the two topics to suggest ways in which we might be able to use knowledge engineering to harness the power of LLMs.
Another type of highly successful presentations linked KO with the real world. Stephann Makri revealed more about serendipity, including a bookshop organized around the idea of serendipitous discovery. It worked, but not for goal-directed searching (which is I guess what you would imagine). Other real-world talks included the “this is what we learned” case study. Martin White’s talk about enterprise search impressed me in that yet again he discovered new things to say about a topic that you thought you were familiar with. So, for example, he revealed that around 5%-10% of all enterprise content comprises PowerPoint slide deck presentations, often with cryptic titles that don’t reveal their contents, so they become a challenge for any discovery system. Another remarkable fact was that 63% of enterprise searching is for people.
A one-off (and unexpected) highlight of the conference was a presentation with immediate practical benefit. David Haynes described an ISKO working group project to explore academic library interfaces, to see if library search interfaces can be improved. It is quite revealing that the need for a critical appraisal comes from the consumers rather than the producers. Anyone who uses these tools can suggest improvements, yet, as Haynes pointed out, library discovery systems are at core a black box: the vendors don’t reveal the principles of their search systems.
All in all, a conference with fascinating moments: not so much an example of knowledge organization as a medley of very disparate talks and agendas.