If you are seeking a sense of tradition, here was where to find it. This year’s Fiesole Retreat was held at the University of Tübingen, founded by Eberhard the Bearded, no less, in 1477, although the Alte Aula, the conference room, from 1547, is by comparison relatively recent. The title “retreat” suggests quiet meditation and contemplation, and this room, with its impressive wooden construction, was certainly an appropriate place to reflect.
On the agenda for the conference were some suitably heavyweight topics to consider: Does reading matter in the age of AI? Who should we trust in scholarly communication? What, indeed, is the role of the library today?
What is a library for?
Richard Ovenden, head of the Oxford Bodleian Library, is in no doubt about libraries’ public role. Of his newspaper journalism, he says, “I no longer write for the library community, but for decision-makers”. His wide-ranging keynote talked about current challenges for all libraries, academic or public, and possible responses. Drawing on thinkers such as John Rawls, he stated “Libraries are the infrastructure of democracy”. Among his recommendations were a call to diversify income for libraries, and pointed out how 10% of the Bodleian turnover now comes from commercial activities — including (unknown to me) using the site for weddings: 73 couples were married at the Bodleian last year.
One of the more unexpected activities of a library might be to examine fanfiction. Sarah Schnitzler, a researcher at the German Literature Archive, talked about fanfiction and its metadata, which is, in effect, a highly individual type of library catalogue.
Benefits of AI
It was heartening to evidence of how AI had benefitted scholarly workflows. Stephen Rhind-Tutt, head of Coherent Digital, revealed how using ML to add metadata had reduced costs dramatically, from $1 per item at his former company Alexander Street Press in 2016, to $0.01 per item at Coherent, while only reducing accuracy from 99.9% to 95%. Coherent’s existence, in other words, needed such a gain in metadata efficiency for it to be viable. Gaetan Rappo described how he uses AI on thousands of early Japanese Buddhist texts to help indicate borrowings and sources – an impressive example of TDM (text and data mining) on a corpus that would not have seemed a suitable candidate for AI.
Do we still need to read?
The rise of genAI has led to questions about whether we need to read today. Marjorie Levine-Clark, a professor of history, described her course in Victorian Studies to students at the University of Colorado, Denver, and revealed some remarkable findings. Apparently, only 20%-30% of students complete reading assignments. Her statement “The long read is no longer a viable default assumption in undergraduate history course design” sounded bleak, but it’s worth remembering that students have been finding ways to avoid reading texts well before AI came along, by using study guides and encyclopedias. Nonetheless, showed how, by the end of her course, her students were reading, and engaged with, classic Victorian novels – and using AI intelligently to compare responses.
One of the most fascinating sessions was a report from a team at Tübingen on the effect of AI on scholarly reading. The project tracked what they called “hybrid epistemic practices”, in other words, using AI as one component of scholarly reading. At one extreme, a student admitted to only using the AI, but most of the 500 interviews with students suggested an “assemblage” of AI-based and human reading. The presenter, Lukas Griessi, was hesitant to praise or to condemn such practices, but clearly, this project is providing evidence of actual practice that is sorely needed at a time when many authors offer deeply pessimistic forecasts for reading.
Projects at Tübingen
Alongside these major themes, Karla Pollmann, President of the University of Tübingen, took advantage of the event to showcase a number of relevant projects taking place at the university, including the University Biobank, which inverted your assumptions about a data repository (“an empty biobank is a good biobank”), plus the ZOERR team, dedicated to producing Open Educational Resources, and a digital project to capture Asian mantras. Quite apart from presenting to the conference, more than one of these project teams noted how they themselves had learned from the presentations about what is happening at their own institution.
Conclusion
As Jim O’Donnell pointed out in his concluding remarks, libraries were invented when books were scarce: today, we have more books than in Chaucer’s day on our smartphone. If books are no longer scarce, then it is vital that we identify best procedures for accessing and manipulating information, both in science and in humanities. This conference provided some real insights into current, and future, practice.
Leave a Reply