Monday, December 23, 2019

Artificial Intelligence and Libraries

HAL9000 from the movie 2001:
A Space Odyssey.  (Image from
Wikimedia Commons)

artificial intelligence:  a branch of computer science dealing with the simulation of intelligent behavior in computers (Merriam Webster)

bias:  an inclination of temperament or outlook especially : a personal and sometimes unreasoned judgment : PREJUDICE (Merriam Webster)

Back in early November I attended the Law Librarians of New England (LLNE) fall 2019 meeting that focused on artificial intelligence (AI) and algorithms in law libraries and legal practice.  AI is and will continue to be a hot topic with the development of “deep fakes,” facial recognition software, Alexa and Siri, targeted advertisement, and self-driving cars.  But how does AI affect libraries and their patrons?

Google’s algorithms use the billions of searches users perform on a daily
bases as data and learn from them in order to predict and/or suggest future searches.

AI has revolutionized both the ways in which we seek information and the speed at which we receive it.  When you use a database or search engine, however, you most likely aren’t thinking about the algorithm that is being used to retrieve results for you, who created the algorithm, and what kind of data it’s drawing from.  I left the fall meeting realizing that everyone should be asking these types questions.  There’s a dangerous expectation that computer and web-based programs are designed to give the user impartial results that represent the truth; however, what if an algorithm intentionally or implicitly reflects the bias of its creator and/or the biases fed into it by its millions of users?  University of Southern California communications professor Safiya Umoja Noble, for her book Algorithms of Oppression, spent 6 years researching Google’s search algorithms and found that the tool not only tends to reflect the values of white western men, who also make up the majority of its builders, but also fails to represent--and even perpetuates negative stereotypes of--minorities and women.  This is a problem that extends far beyond Google, which is why it’s important to consider the objectivity of the pipeline that is delivering information to you.

Another issue with AI is the lack of standardization and transparency.  Librarians love standardization, which is why there are rules for metadata creation that cataloging librarians across the world follow to create access to library materials (for example: MARC records and Library of Congress subject headings).  Library catalog records are usually completely transparent and users can see the “code” used to index an item.


Proprietary companies understandably do not want to share the secret inner workings of their products, but this poses a problem when the user wants to better understand why one database pulls X results while another database pulls Y results, or why the results are listed in a certain order, and most importantly where this data is coming from and if its complete.

Admittedly, I myself still know so little about AI and there are so many different facets and ethical and privacy issues to consider, but what I do know is that it’s important for library users from all backgrounds and disciplines to be aware of the repercussions of too much reliance on it when seeking information.  DON’T assume that AI is smarter than you, DO assume the probability of bias, DON’T assume that it draws from an exhaustive bank of quality data, and DON’T assume that having difficulty finding something in a search engine or database means that it doesn’t exist.

Kaitlin Connolly
Reference Department


Further reading: