In the discipline of artificial intelligence (AI) known as machine learning (ML), “supervised learning” takes place when a set of data that has already been classified by humans is used to “teach” the machine how to classify additional data on its own. While it is underpinned by a lot of math and computer science, it is essentially a way of teaching machines through the use of examples.
Similarly, if one were to offer an optimal strategy for librarianship when it comes to AI/ML technology, they might build on examples of past technological disruptions, see how librarians’ reactions varied, and assess how outcomes of those approaches varied. I have to point out that the examples that I will use below are from my experience working as a library professional for over twenty years, but to be transparent, the dataset is limited and is biased (coming only from my own perspective), so the conclusions will be debatable. Still, I offer them as part of the ongoing conversation about AI/ML in libraries that was generated in the IDEA (Innovation, Disruption, Enquiry, Access) Institute in July 2022.
I began working as a library professional in the 1990s and remember the field going through a number of technological changes. As the internet began to gain purchase in the public imagination there was much discussion about libraries and librarians no longer being needed because the internet would provide the world’s knowledge to its users. Some librarians in that era took on an adversarial us-versus-them mentality. I remember similar responses to the growing popularity of a new search engine called Google. The approach of the resistors can be characterized as those who chose to remain aloof and ignorant of the technological development, those who were disdainful of it (particularly if they could point to an example of the technology having failed), and those who displayed a more active antagonism toward the technology in question, dismissing any opportunities that might have arisen from the use of that technology. Resistance to technological disruption was futile.
The libraries and librarians who were most successful in a world disrupted by technology did so by leveraging these technologies in ways that furthered their mission. They did so by being open enough to learn what might be possible (experimenting) and being mindful enough to not to simply chase “the shiny,” heedless of potential downsides. They recognized that a one-time experiment that may have “failed” did not make the technology a failure or the endeavor a waste of resources. Instead, they watched how things evolved, they reexamined their initial implementation and processes, and pivoted as needed. They learned and grew. Ultimately, librarians who used these tools to increase their own literacy and efficiency (and the literacy and efficiency of their users) were empowered by the technologies in question.
Librarianship embraces a strong human-centered ethos and critical thinking. Librarians help users grow their information/digital literacy so that they are empowered in an information systems-driven world. For these reasons the AI/ML world, which is often pushed forward by ethically questionable profit-making endeavors and developed by practitioners with little diversity in identity or perspective, needs librarians. In the development of the powerful AI/ML technologies that are now core to our daily lives, biases are coded into the tools and datasets used. So, not only do librarians need to grapple with AI/ML, AI/ML would be well-served by engagement with librarianship.
So, what strategies can I suggest for libraries and librarians in adjusting to an AI/ML-reliant world, while ensuring that the ideals of librarianship are grounding this work?
– Grow our technological (AI/ML) literacy:
– Continue to learn about developments in AI/ML broadly and where they are being used by various sectors of the economy and in the government, while paying attention to issues of bias and implications of the power structures involved
– Continue to discuss the deeper ethical implications of AI/ML as it is developed and deployed
– Build infrastructure – prepare library outputs – data (including the use of our collections as data) – in order to participate mindfully in an AI/ML-driven world with an eye toward equity, diversity, inclusion, and accessibility (EDIA) and attention to privacy concerns
– Learn from the technology by building and experimenting with it, but only deploying AI/ML projects if they meet the requirement that any technology to be adopted must serve people not the other way around (centering the EDIA and privacy/surveillance issues, for example)
– Leverage the technology in whatever ways we can to improve users’ experience of the library, librarians’ capacity to meet demand when it comes to research, metadata / digitization, and and to increase efficiencies in research processes and imprand other librarian functions
In the final IDEA Institute Showcase of presentations, these strategies are being used. The projects range in the specifics of their outcomes, be they oriented toward the improvement of user services (e.g., reference chatbots), improving research and discovery services, or improving tools for digitization and analysis of repository materials. But in all cases, IDEA fellows’ projects are building solutions that serve the mission of librarianship, while being informed by a professional ethos of human-centered design, in which EDIA and privacy/surveillance concerns are centered.