Design for AI in Libraries and Archives at the Fantastic Futures Conference

December 18, 2018
Catherine Nicole Coleman
Group photo of workshop participants
On December 4, 2018 in Oslo we held a day-long "Design for AI in Libraries and Archives” workshop in collaboration with the National Library of Norway the day before the Fantastic Futures Conference. We set out to build a shared understanding of the possibilities and practical applications of library-inspired artificial intelligence. There were overarching concerns expressed about how copyright restrictions will limit what we can do with content and whether libraries will have the capability to carry out the work that needs to be done to operationalize content. Though neither of those concerns were addressed directly in the workshop, we agreed that fostering AI literacy across an organization will help generate ideas and enthusiasm. We also agreed to continue the discussion. (See the Google group ai4lib and the ai4GLAM Slack workspace)
 
These topic threads ran through the workgroup discussions:
 
Metadata
Some suggested that top-down hierarchical classification schemes will no longer be needed at all! Others suggested changes to how metadata is created. It could be crowd-sourced, it should be iterative, it may be an opportunity for human-machine collaboration where unsupervised clustering is part of a dimensionality reduction leading to further iterative classification evaluated by humans along the way. There was also excitement about the possibility of generating new types of metadata, like metadata that will provide temporal context. 
 
Discovery and Non-verbal search
The group found that in our data-rich world there is a shift to exploring more than searching. The hope was that we could use a combination of AI techniques to improve the quality of results, perhaps even make results more personalized. For both text-based search and non-verbal search there was interest in multiple modes of discovery that include precise matching, similarity, and degrees or types of relatedness. One vector explored was using stylometrics to reveal features and qualities of a text rather than just searching for matching text strings. There was quite a bit of interest in the opportunity for discovery based on the content itself rather than descriptive metadata, whether artificially generated or human generated. 
 
Interface and APIs
Though interface was not an identified discussion topic, several of the groups recognized that making AI usable will require designing new user interfaces. The terms used to describe these imagined new interfaces include: intuitive, dynamic, responsive, iterative, interactive. People want some forms of analysis to be integral to the discovery process. And there seemed to be consensus that, consistent with the notion of non-verbal search, an interface should be able to take various forms of input including voice, images, recorded audio, etc. The interaction desired took the form of engaged feedback and the ability to iteratively refine a query through a process of question and result.
 
Bias
With so many warnings circulating about the willingness to blindly trust algorithms as unbiased, or capable of somehow correcting human bias, it was nice to be with a group of librarians who know all too well that bias is unavoidable, is pervasive, is human, and needs to be examined.  This group tended to be concerned with how, by looking more deeply and broadly at content in context, we could more systematically uncover and draw attention to bias in our collections.
 
Interspersed between the various group activities, we watched brief videos on AI-related topics from the general societal impact of the technology to very specific applications relevant to cultural heritage.
 

Kevin Kelly: How AI can bring on a second industrial revolution

Jean-Baptise Michel and Erez Lieberman Aiden: What we learned from 5 million books

Gregory Heyworth: Discovering the secrets of ancient texts

Maurice Conti: The Incredible Inventions of Intuitive AI   

 

AI Workshop Participants in photo above:

Back row: Frank Busse, Thomas Edvardsen, Thomas Langvann, Freddy Wetjen, Roger Mathisen, Peter Leinen, Annie Murray, Jessie Keck, Tom Cramer, Sonia Wronkowska, Jeff Steward, Andre Walsøe. Middle Row: Uldis Bojars, Adam Metallo, Renaldas Gudauskas, Christian Boesgaard, Stace Maples, Svein Arne Brygfjeld, Nina Hyvönen, Stu Snydman. Front Row: Eric Lease Morgan, Tonny Skovgård Jensen, , Andrea Gasparini, Nicole Coleman, Michele Casalini. (Not pictured: Philip Schreur.)

 

accessibilityaccessprivsarrow-circle-rightaskus-chataskus-librarianbarsblogsclosecoffeecomputercomputersulcontactsconversationcopierelectricaloutleteventsexternal-linkfacebook-circlegroupstudyhoursindividualinterlibrarynewsnextoffcampusopenlateoutdoorpeoplepolicypreviousprinterprojectsquietreservesscannersearchstudysupportingtabletourstwitter-circleworking