Managing Bias When Library Collections Become Data

Everett-Hayes, Lauren

Coleman, C. N. (2020). Managing bias when library collections become data. International Journal of Librarianship, 5(1), 8-19. https://doi.org/10.23974/ijol.2020.vol5.1.162

Summary

Catherine Coleman’s article addresses AI developments and how libraries should be critical of their integration into library functions, instead focusing on the ethos of what libraries do for patrons. For one, the point is brought up that AI has shown to be biased because of the nature of how it gets its data. It can only take what it can access, which is not a complete view of the information someone may try to access. The author illustrates that a collection is where library data is concentrated, and, when considering how to incorporate AI, how you access information becomes extremely more relevant (Coleman, 2020).

A core thing the author seemed to want to get across was that AI should be wielded as a tool instead of as a solution as it could help with assessing bias in collections. A current example of overarching bias are the paradoxical LOC subject headings in how they are necessary for categorizing but also cause problems through misrepresentation, racism, etc. The author also expertly speaks about how libraries are more relevant than ever with the integration of AI:

At this moment when there are as many papers about the successes of AI research as there are papers calling out algorithmic bias, data bias, and setting forth principles of AI practice, libraries need to do much more than provide curated data to AI researchers. Libraries need to apply the principles of the profession to managing bias in AI-based systems. (Coleman, 2020, p. 16)

Lastly, Coleman (2020) sums up the article perfectly with the following quote, which pairs well with her call to action that is illustrated throughout: “Libraries need what AI has to offer, but AI needs what librarians have to offer even more” (p. 16).

Evaluation

Overall, I found this article well articulated and felt that it pushed the subject on AI in libraries in effective ways that both explored how AI could help and where it has limitations. Unsurprisingly, there are a lot of things librarians have to consider regarding our library Code of Ethics, copyright, and accessibility to our collections as there is more of a push to use AI and allow it to access our collections without a leash. However, this article talks about how AI can be both a barrier and an asset that continuously needs human input and discretion. In this back-and-forth, there ends up being a lot of really great questions and points made by Coleman that gives us a lot to think about and prepare for, but also feel empowered by. I don't think we have to be afraid of AI replacing our work or making libraries obsolete. The way in which humans understand other humans and what they need will always exist through the nature of research and making connections to the scholarly conversations out there. AI might get robust enough to aid us in seeing our biases and analyzing collections on a deeper level, but I don't think we will use it to put the books on the shelf, since human patrons will always be our focus.

As an honorable mention that might be interesting to others, there is an open access book that is mentioned in this article that gives a humanistic perspective on data that I would like to note here: All Data Are Local: Thinking Critically in a Data-Driven Society