Columns

XXVIII.3 May - June 2021
Page: 18
Digital Citation

A lifelong journey: Addressing racial biases in design and research processes


Authors:
Daria Loi

back to top 

One of many pivotal events in 2020 was the global wave of civil unrest and protests that began after the killing of George Floyd. Captured on video, that moment of despicable brutality, when a police officer knelt on the neck of the 46-year-old African-American man for 8 minutes and 46 seconds, enabled many to see what many others already knew: Racism exists and persists to the detriment of the fabric of our society. And racism, as well as racial biases, permeates our everyday lives, actions, decisions, and behaviors.

Much unfolded after that, including global tributes to honor the lives of Mr. Floyd and many others, such as Breonna Taylor, the Louisville, Kentucky, EMT who was shot to death by police during a botched raid on her apartment two months before Floyd was killed. There were protests and demonstrations over systemic racism within law enforcement and beyond; acts and declarations of solidarity; charitable donations to BIPOC (Black, Indigenous, and people of color) communities; increasingly proactive discussions about everyday racism and discrimination; and the removal of racially insensitive or inappropriate characters. Many individuals and organizations committed to embark on a lifelong journey focused on debiasing their days, lives, practices, and work—this column reports on the beginning of one such journey.

In 2019, Obermeyer et al. [1] demonstrated evidence of racial biases in a widely used algorithm by a leading health services company. The technology, by assigning the same level of risk when Black patients were sicker than white patients, reduced the number of Black patients identified for extra care by more than half. This occurred as the algorithm falsely concluded that Black patients are healthier than equally sick white patients, as less money was spent on Black patients who had the same level of need. The algorithm used health costs as a proxy for health needs. The algorithm was unintentionally racist and, as concluded the researchers, the remedy of such a disparity would increase the percentage of Black patients receiving additional help from 17.7 to 46.5 percent—a substantial difference.

Like the example discussed by Obermeyer et al., the history of racist and racially biased products and brands is long and paved with depressingly numerous case studies. In the 1970s, Kodak coated its film to favor white skin tones [2]. The past decade alone holds notable examples: Google Photos incorrectly identified and labeled two African Americans as "gorillas" [3]; Microsoft's chatbot Tay learned racist language—in under 24 hours—when fed by Internet trolls taking advantage of its machine learning algorithms [3]; Amazon's Rekognition system, according to multiple tests, falsely identified darker-skinned females about 30 percent of the time and falsely matched people of color to arrest photos at a higher rate [3]; and Qiaobi's detergent was advertised by portraying an Asian woman placing a detergent pod into the mouth of a Black worker who, after being pushed into a washing machine, emerged as a pale Asian man [4].


Racist and racially biased products mirror the racist and racially biased cultures, processes, companies, and individuals that created them.


In the past few years, in response to organizational introspection (or simply forced by customers' outrage or sponsors' pressure), a number of companies started reviewing their brand identity and product offerings, from the NFL's Washington football team changing their reprehensible former name to PepsiCo retiring its Aunt Jemima brand [5]. That said, the journey has merely started; undoubtedly much work is still needed, globally and across multiple vectors.

Racist and racially biased products mirror the racist and racially biased cultures, processes, companies, and individuals that created them. And for every blatantly racist product, there are thousands built on top of (or through) racial microaggressions [6]—and while this term is used in literature, micro implies that the aggression or its effect is small, which is inaccurate. These subtle yet equally toxic processes are antithetical to human-centered design: They exclude and damage fellow humans and, ultimately, our society.

In 2007, Sue et al. [6] discussed how microaggressions appear in three forms: microassault, microinsult, and microinvalidation. Microassaults [6] are "conscious and intentional actions or slurs" (e.g., racial epithets, display of swastikas, deliberately serving a white person before a person of color in a restaurant). Microinsults [6] are "verbal and nonverbal communications that subtly convey rudeness and insensitivity and demean a person's racial heritage or identity" (e.g., asking a colleague of color How did you get your job?, implying they "may have landed it through an affirmative action or quota system"). Microinvalidations [6] are "communications that subtly exclude, negate, or nullify the thoughts, feelings, or experiential reality of a person of color" (e.g., asking an Asian American where they were born, "conveying the message that they are perpetual foreigners in their own land").

ins01.gif

Much literature provides insights on how and where microaggressions are embodied by the products, services, systems, and processes that surround us. While painful to admit, the reality is that much of what surrounds us is a vehicle for daily microassaults, microinsults, and microinvalidations. In this column I rely on existing evidence and literature to state what we all know very well: We are all responsible and must all assume responsibility.

So what can HCI thinkers, practitioners, and decision makers do? Turns out, a lot.

In 2020 I had the privilege of leading, together with a number of esteemed colleagues at Mozilla, an effort focused on driving representation in research and design efforts based on race and ethnicity. The effort included evaluations of design and research processes through a series of participatory workshops with internal stakeholders, a dedicated core team in charge of analyzing all data, and sponsoring executives committed to the effort. During that time we all had the opportunity to be exposed to numerous resources and expert presentations, as well as training to deepen our understanding of how racial biases make their way into one's business practices and outcomes.

After a number of months, five key recommendations were developed, approved by senior executives and shared across the company, with the commitment of executing them in 2021. Below I share the five steps that we identified and the journey we are beginning.

First and foremost, for products to be debiased and to cater to the needs of all types of humans, the companies that develop those products must be representative of the diverse communities they pledge to serve. This requires the active inclusion of diverse talent, points of view, and voices, as well as a commitment to address gaps in intersectional BIPOC communities' representation.

In big tech, for instance, "the number of black designers practicing in 2019 was a mere 3% of the industry total" and overall "there have been little to no gains in the last five years for black employees" [3]. To design and develop products that truly cater to all types of humans and all types of global markets, companies need to urgently and substantially shift these numbers. This can be tackled by creating, delivering, resourcing, and nurturing hiring plans focused on addressing the underrepresentation of intersectional BIPOC communities. For instance: Promote roles and recruit in racially diverse communities; value broad and relevant lived and professional experience; increase the percentage of racially diverse talent throughout the acquisition pipeline; and publicly share journeys and learnings. These hiring plans need to continuously grow, improve, and be consistent over time. As Ben Hecht writes in the Harvard Business Review, "Achieving racial equity in the workplace will be one of the most important issues that companies will tackle in the coming decade" [7].

A second step to consider is to create checks and balances for all research, design, and development processes. Process checkpoints are key to ensure that racial inclusivity is present throughout a product's design and development lifecycle. These not only require designers, researchers, engineers, and product managers to align on timelines, targets, and goals as they pertain to racial diversity, but also to debrief on progress over time. The purpose of checks and balances is to build behavior and habits that integrate diversity and inclusion throughout business processes.

Additionally, ad hoc processes and checklists to review for racial bias can be used when developing research scripts. Tools used for recruiting and sampling research participants can be used to oversample racially marginalized communities to ensure sufficient understanding and representation in one's results. Desk research should also be inclusive in nature (e.g., BIPOC authors and racially inclusive methods).

A third step relates to third parties and vendors that many organizations rely on to complete key activities and tasks, from participant recruitment to research and design. During our analysis, we have observed that biases can creep into one's processes through that route.

At a basic level, if a third party or vendor as an organization is not diverse, then it is likely that their lack of diversity will influence how they plan and conduct their business, with biases affecting their processes, decisions, focus areas, and priorities. This can be addressed by establishing objective vendor requirements to cultivate intentionally diverse insights and processes, and to evaluate vendors against such criteria.

Moreover, if the participants and panels of external recruiters lack diversity, then research efforts will lack diverse perspectives, impacting one's ability to make unbiased product decisions. To address this, a strategy is to identify partners and tools that have race as criteria for participant selection. An alternative route is to identify partners that specialize in BIPOC audiences. In addition, one can develop or identify a racially diverse panel on retainer to rely upon for quick day-to-day research iterations.

We have also learned that some vendors tend to screen out participants based on their English competency, fearing that their clients will reject their bids due to the implied need for costly translators. To tackle this, it is key to allocate budget, resources, and time to include translators for English as a Second Language (ESL) speakers who participate in research.

A fourth step relates to evaluating our practices to identify the organization's knowledge and depth in relation to BIPOC communities. In our case, we flagged gaps in our ability and knowledge portfolio to identify product opportunities, use cases, trends, and insights that include, serve, and address the needs of BIPOC populations. While most organizations have a wealth of knowledge about their customers, major gaps often exist in their foundational understanding of BIPOC populations. The reality is that the vast majority of companies and organizations have much to learn. One could consider conducting foundational research focused on BIPOC populations to identify product gaps, use cases, rich user trends, and other insights. In addition to foundational research, one can dedicate ongoing research efforts and funding to the needs of BIPOC communities. Ultimately, it is key to build a knowledge library on the needs, experiences, and behaviors of BIPOC communities that is intrinsically included across one's product exploration and design processes.


While most organizations have a wealth of knowledge about their customers, major gaps often exist in their understanding of BIPOC populations.


A fifth step relates to market focus. Through our research, we have identified that in a number of cases (and ways), market choices impact racial inclusivity. In some cases, this is a consequence of the market one chooses to target (e.g., research focused on markets only in the U.S., Canada, and Western Europe due to resource allocation). In other cases, racial inclusivity is impacted as a consequence of insufficient diversity within a selected market; for instance, activities related to a product for the U.S. market may leverage research that is not representative of U.S. populations. In still other instances, work to serve a target population may, as a consequence, have the potential to harm those who are not part of that target population.

To address these issues, it is key to begin each project by stating one's choice of target market and to document how such choices impact racial inclusivity. It is imperative to have a clear focus on diversity within selected markets and to examine as well as articulate risks where a project may harm people outside of the focus area. Of course, it is fundamental that organizations recognize the importance of expanding their choice of target market to include traditionally underserved communities or populations.

Above I shared five steps that my colleagues and I identified as the beginnings of a lifelong journey focused on driving representation in research and design efforts based on race and ethnicity. Organizations and teams interested in this journey should consider engaging in work targeted to 1) addressing gaps in intersectional BIPOC communities' representation; 2) creating checks and balances for all research, design, and development processes; 3) evaluating and addressing biases related to third parties and vendors; 4) acquiring knowledge and depth in relation to BIPOC communities to identify product opportunities, use cases, trends, and insights that include, serve, and address their needs; and 5) being deliberate and having a clear focus on diversity within selected target markets, articulating risks where a project may harm people outside of the focus area.

ins02.gif

2020 shed light yet again on the indisputable reality that racism and racial biases permeate our everyday lives, actions, decisions, and behaviors. It was a year that asked us to see the ways in which we are all responsible and must all assume responsibility. Addressing biases in design and research is one of the many opportunities that HCI practitioners, thinkers, and decision makers have to benefit fellow humans, communities, and society. It is also a key way to honor and be truthful to the human-centered processes at the heart of our HCI practice.

back to top  Acknowledgments

I wish to acknowledge Mozilla colleagues who participated in the effort discussed in this column, in particular James Keller, Leslie Gray, Sharon Bautista, Steven Potter, and Thomas Lodato. The work I shared would not have occurred without their knowledgeable input and relentless dedication. I also wish to thank Adam Seligman and David Camp, executive sponsors and ongoing supporters of the effort.

back to top  References

1. Obermeyer, Z., Powers, B., Vogeli, C., and Mullainathan, S. Dissecting racial bias in an algorithm used to manage the health of populations. Science 366, 6464 (2019), 447–453. DOI: 10.1126/science.aax2342

2. Lewis, S. The racial bias built into photography. The New York Times. Apr. 25, 2019; https://www.nytimes.com/2019/04/25/lens/sarah-lewis-racial-bias-photography.html

3. Palaniyappan, A. A brief history of how racism manifests itself in design and how we can learn from it. UX Planet. Jun. 4, 2020; https://uxplanet.org/a-brief-history-of-how-racism-manifests-itself-in-design-and-how-we-can-learn-from-it-141b1b5ddd4b

4. Bromwich, J.E. Chinese detergent ad draws charges of racism. The New York Times. May 27, 2016; https://www.nytimes.com/2016/05/28/world/asia/chinese-detergent-ad-race-qiaobi.html

5. Kesslen, B. Aunt Jemima brand to change name, remove image that Quaker says is 'based on a racial stereotype.' NBCNews. Jun. 17, 2020; https://www.nbcnews.com/news/us-news/aunt-jemima-brand-will-change-name-remove-image-quaker-says-n1231260

6. Sue, D.W., Capodilupo, C.M., Torino, G.C., Bucceri, J.M., Holder, A.M., Nadal, K.L., and Esquilin, M. Racial microaggressions in everyday life: Implications for clinical practice. Am Psychol. 62, 4 (May–Jun., 2007), 271–86. DOI: 10.1037/0003-066X.62.4.271. PMID: 17516773

7. Hecht, B. Moving beyond diversity toward racial equity. Harvard Business Review. Jun. 16, 2020; https://hbr.org/2020/06/moving-beyond-diversity-toward-racial-equity

back to top  Author

Daria Loi heads Design & People Experiences at Mozilla. Her work revolves around mixing design strategy with UX research to enrich people's everyday lives. In 2018 she was recognized as one of Italy's 50 most inspiring women in tech. dloi@mozilla.com

back to top 

Copyright held by author

The Digital Library is published by the Association for Computing Machinery. Copyright © 2021 ACM, Inc.

Post Comment


No Comments Found