Algorithmic Bias

My Bias Isn't the Only Obstacle to Comprehensive Research

by Jackie Whiting

I first encountered the idea that bias can be coded into the algorithms that run the platforms I use on a daily basis when reading John Maeda's book How to Speak Machine. Maeda discusses disparities in hiring practices that result in a disproportionate number of what he calls "pale males" throughout the tech sector, particularly in managerial and executive positions. He considers the impact the situation has on interpersonal relations and corporate culture as well as on the computer code that these technicians write. More recently I discovered Algorithms of Oppression by Safiya Noble, who picks up where Maeda left off and delivers a scathing examination of the ways in which the bias coded into Internet search engines returns stereotypical at best and downright offensive at worst search results. In fact, the algorithms themselves, through the autocomplete feature, steer searchers to these narrow-minded and prejudiced results.

Think of the impact on information acquisition by us, our colleagues, our students! As search algorithms become increasingly responsive to search criteria that are phrased as questions, this issue deepens. Of course, we teach our students to compile lists of keywords and to constantly extend and add to that list as they accumulate more information and encounter multiple perspectives on the subject of their research. We stress the importance of using Boolean operators to refine their results. Still, when they (and sometimes our colleagues—and maybe even sometimes we) open a browser and begin searching, into the omni bar we type the question we are asking.

As awareness of this issue has been raised, the search team at Google, for example, has eliminated the auto-complete function for phrases such as "why do black girls…", which was one of the first search instances that got Noble's attention and precipitated her research (p.17). This re-coding work was underway at the time of Noble's writing and is also reflected in changes to search for other groups of people as well. Their search refinement work includes an "inclusive images competition" to populate the databases from which image search draws with more diverse, inclusive, and authentic representations. Nonetheless, as Maeda reminds us: "It wasn't your fault that you didn't know much about it, as you couldn't have noticed it in the first place. But now you know it's there, and everywhere" (p. 174).

Hold on a second if you think that sending students to your meticulously curated collection will either avoid or help to remedy this situation. Noble dedicates a segment of her book to an examination of the Library of Congress Subject Headings and the Dewey Decimal System, and the similarities in those classifications to the algorithms that control search are distressing. Noble comments: "Those who have the power to design systems—classification or technical—hold the ability to prioritize hierarchical schemes that privilege certain types of information over others…Information organization is a matter of sociopolitical and historical processes that serve particular interests" (p.139). She recounts the ongoing revisions of labels in an effort to remove offensive designations such as "Yellow Peril" and still laments that many headings and designations still carry a reference to being a departure from white maleness. "Women Accountants" by its very existence as a subject heading implies that accountants, by design or default, are male.

Momentum is growing to examine subject headings and access to information. In fact, a group of students from Dartmouth College organized a movement to petition the Library of Congress to change the subject heading "Illegal Alien." Lourdes Gutierrez Najerra, an assistant professor at Western Washington University and former advisor to the Dartmouth students, said: "It's not just a catalog word. It means that you are giving value and you are giving credence to that subject heading, and that is the way you are cataloging knowledge on a global scale…the fact that it has been used and codified in the subject heading has tremendous power because that shapes the way we think and write." Her message is personalized by Melissa Padilla, Dartmouth class of 2017 and one of the student organizers, who commented: "to see a word that very starkly represents one way to dehumanize undocumented people emotionally hurt me but did make me also not really trust the library as a system." Their story is told in the documentary, Change the Subject, available on the Dartmouth Library's YouTube channel. Dartmouth College and Dartmouth Library are not alone in doing this work. In fact, the X_wi7x_wa Library at the University of British Columbia is a center for "academic and community Indigenous scholarship. Its collections and services reflect Aboriginal approaches to teaching, learning, and research" (https://xwi7xwa.library.ubc.ca/) Their library research guides provide resources and strategies to facilitate de-colonial and anti-racist research.

All of this matters because the point of organizing information is to make it discoverable and available and valuable to everyone. Consider, for example, something like a shared team Google Drive or some other digital platform for document storage and collaboration. I sometimes find these drives frustrating because I am subject to the organizational structures (or lack thereof) used by the other members. I have to learn someone else's naming conventions and categorization (or they have to adapt to mine). What I can access with ease is dependent on being able to navigate how the materials have been organized and that system is the result of the priorities of the person in charge of the space.

Accessibility of information is the underpinning of Google as a search engine. Their mission statement declares: "Our company mission is to organize the world's information and make it universally accessible and useful. That's why Search makes it easy to discover a broad range of information from a wide variety of sources." And Carla Hayden, the first woman and first Black person to lead the Library of Congress, in her open letter on the library's website says: "The Library preserves and provides access to a rich, diverse and enduring source of knowledge to inform, inspire and engage you in your intellectual and creative endeavors. Whether you are new to the Library of Congress or an experienced researcher, we have a world-class staff ready to assist you online and in person" (https://www.loc.gov/about/). Our information access is their declared priority. The challenges these organizers of information are undertaking to create equitable and inclusive systems is a herculean one. As these structures continue to shift and be rebuilt, our literacy around how information is organized and accessed is essential to our understanding of the world we (re)search.