At least in theory, algorithms based on machine-learning remove human biases from decisions by removing humans from the equation. But in reality, biases end up baked into outcomes, because humans create the algorithms upon which machine learning is based.

There are numerous accounts, for example, of facial-recognition software and object-detection technology misidentifying dark-skinned people. One report from Georgia Tech researchers even posits that dark-skinned pedestrians could be more likely to be hit by self-driving cars. And a recent ProPublica report found that algorithms used to predict which criminals are most likely to re-offend often assessed Black people at higher risk than white people, even if the white person committed a more serious crime.

Orakwue Arinze, professor of decision sciences and management information systems (MIS) in the LeBow College of Business, is working on a project to prevent and mitigate racial bias in machines by modifying the Cross-Industry Standard Process for Data Mining (CRISP-DM), a popular methodology for data analytics and machine-learning development. The goal is to make modifications that are easy to understand, practical and demonstrably effective in helping developers de-bias their applications.

He chose to work with something already in use because it doesn’t require creating a new framework. Instead, it adapts an already established framework to take the biases — and biased decisions they produce — out of the equation.

“While it is possible to create a brand-new methodology for developing machine-learning systems, piggybacking on the popular CRISP-DM methodology increases the probability of adoption by data scientists,” he says.

In 1936, victor hugo Green wrote The Negro Motorist Green Book, which became an annual guide book for Black road trippers directing them toward hotels, restaurants, repair shops and gas stations that were safe for Black travelers, and spotlighting which towns to avoid.

Brea M. Heidelberg, associate professor and program director in the Entertainment & Arts Management program in the Westphal College of Media Arts & Design, argues that a sort of “Intellectual Green Book” exists in the form of informal networks of guidance intended to combat the intersectional oppression that Black women and Black non-binary people face in the nonprofit sector. She is undertaking an exploratory study to demonstrate its existence and to begin investigating how it is created, maintained and used.

Organizations like the Arts Administrators of Color Network and Women of Color in the Arts already work on these issues “but the process and impact hasn’t been theorized and researched before,” Heidelberg says. “My research is based upon the premise that this specific type of green book exists, but that it hasn’t been fully articulated yet.”

The project seeks to answer questions such as, how do Black women and Black non-binary individuals in the nonprofit arts make career choices? Does one aspect of their identity take precedence when making these decisions? What types of information do they get prior to making that decision? How do they cope with workplace trauma? How do Black women and Black non-binary individuals in the nonprofit sector use internalized gatekeeping tactics on others?

“The Intellectual Green Book may be preventing Black women and nonbinary individuals from leaving the cultural workforce by providing a place to process, heal from and learn how to combat workplace violence,” she says.

Stark racial disparities in the juvenile justice system are already well known. Nonwhite children tend to receive harsher treatment and outcomes in juvenile court. But what mechanisms create these disparities?

One potential factor: school discipline, which could have life-long implications.

“Research shows that being arrested or detained during youth or adolescence can harm prosocial development and may increase the likelihood of repeated crime and punishment into adulthood,” says Kathleen Powell, a postdoctoral fellow in Criminology and Justice Studies in the College of Arts and Sciences. “Identifying the link between school suspensions and justice system involvement for youth and adolescents … is thus critically important to understanding how the justice system can contribute to social inequalities.”

The project looks at how correlations between suspension and the justice system vary by race, and in doing so, asks how a child’s self-identified race or ethnicity moderates the relationship between being suspended and arrested or receiving a criminal sentence.

Powell’s analysis will use quantitative methods and NLSY97 data, a longitudinal survey collected by the U.S. Bureau of Labor Statistics that includes a sample of participants who were ages 12 to 16 as of Dec. 31, 1996, and were interviewed annually between 2007 and 2011 and biannually to the present.

“Not only is the sample large and national in scope, but the data also capture the experiences of a cohort of youth coming of age during a historical moment in which juvenile justice and school discipline policies became increasingly punitive and zero-tolerance, respectively, in nature,” Powell says.

As activists put forth new calls for police reform, Associate Professor of Community Health and Prevention Amy Carroll-Scott and Jennifer Kolker, clinical professor and associate dean for public health practice and external relations, both in the Dornsife School of Public Health and the Urban Health Collaborative, decided to focus their already successful policy surveillance study on examining law enforcement reform efforts in U.S. cities.

“Since George Floyd’s death, the nation has been more supportive of efforts to better understand and develop solutions to prevent these fatal encounters between police and Black individuals (and other residents of color),” says Carroll-Scott. “What we need to know is whether calls for reform are actually resulting in reform. And if so, what are best practices that can be adopted across police systems.”

Between 2017 and 2019, Carroll-Scott and Kolker’s team at the Policy & Community Engagement Core of the Drexel Urban Health Collaborative did similar policy surveillance on seven cities, including Philadelphia. Researchers looked for city policy solutions related to the social determinants of health (i.e., housing, education, the environment) to support cross-sector health-equity strategies, and created cross-city policy briefs that were distributed in the fall.

This new study replicates that existing methodology to identify and monitor policies emerging between May 2020 and May 2021, but focused on policies addressing law enforcement reform. Preliminary findings have revealed policies related to law enforcement transparency and accountability; bans on no-knock warrants; de-escalation tactics in protest situations, pilot police deflection, diversion, or co-responder models; body camera use and footage release; and training programs related to implicit bias and de-escalation protocols. Final results will compare successes and challenges across cities.

Implicit bias flourishes where individuals have the freedom to ignore how their race impacts others as well as the harsh “realities” that others must cope with regularly.

Unfortunately, current multicultural and diversity trainings mostly address explicit bias (i.e. overtly racist acts) while leaving implicit bias intact. They almost always use people of color as the population to be considered, which has the effect of maintaining whiteness as the norm and “pathologizing” the cultural values of non-white groups.

Three College of Nursing and Health Professionals researchers — Ebony White, assistant clinical professor in counseling and family therapy professionals; Kimberly McClellan, assistant clinical professor in undergraduate nursing; and Arun Ramakrishnan, director of research labs — are creating an immersive learning course titled, “Examining Whiteness,” that “flips” that script by instead directing the examination of whiteness, with the help of virtual-reality scenarios.

Immersive learning and virtual-reality technology are key to their approach because immersive experiences have been shown to be effective in reducing implicit bias by encouraging empathy and engagement.

The virtual-reality scenarios they are creating will enable participants to take part in two different ways. In the first scenario, they can simply be an observer to how bias plays out in various situations, while data is captured regarding their reactions. The second will allow them to experience the scene as one of the characters, accelerating understanding of how bias affects different demographics.

When Minneapolis police officer Derek Chauvin kneeled on George Floyd’s neck until he died, three police officers did nothing to stop him.

This is called passive bystandership, and it is such a problem in American policing that many departments have enacted peer-intervention policies requiring officers to intervene when their colleagues make mistakes or engage in serious misconduct. However, even the most robust programs have not proven to be effective.

Robert J. Kane, department head and professor of criminology and justice studies at the College of Arts and Sciences, is conducting a pilot study to develop and test a police officer peer-intervention training that treats police officers like clinicians by using the same type of simulation deployed to teach doctors, nurses, and airline flight crews (though never before practiced on law enforcement).

The seven- to 20-minute simulation immerses students, who will be in their final week of field training, in disorientating hypothetical situations, most of which “will involve a more senior officer escalating the encounter to the point where they use excessive force on the person they’re dealing with,” says Kane. During simulations, they’ll measure the degree to which the students intervene, if at all.

Students will be debriefed by instructors and peers, and a facilitator will guide them through a discussion of what they were thinking and/or feeling in the simulation so they can better understand their actions, and identify strategies needed to work through a similar situation in the future.

While we all feel impacted by the pandemic, statistics show that COVID-19 is disproportionately decimating communities of color. Early in the pandemic, in most of the 33 states (plus the District of Columbia) that reported race and ethnicity as part of their COVID-19 cases and deaths, Black people made up a higher proportion of confirmed cases and deaths relative to their share of total population. In six of the 26 states reporting data for Hispanic people in April 2020, these communities made up a greater share of confirmed COVID-19 cases compared to their share of the total population.

“The underlying, well-known health, social and economic disparities present in communities of color are being magnified by the pandemic,” says Loni Philip Tabb, associate professor of biostatistics in the Dornsife School of Public Health. She is co-principal investigator with Scarlett Bellamy, professor and associate dean for diversity, inclusion and faculty development at Dornsife and Leslie Ain McClure, professor and chair of epidemiology and biostatistics.

The researchers, including recent master’s graduate Kara Beck ’21, are using publicly available county-level data to adjust for a lack of consistency in reporting race and ethnicity related to COVID-19 health outcomes. They hope their work will close the gap, for the sake of future researchers, of peer-reviewed, scientifically rigorous evidence that definitively assesses these relationships and how they vary across the country.

Disparities in policing aren’t based on race alone, and yet there’s a dearth of work examining law enforcement through an intersectional lens, says Jason Orne, assistant professor of sociology in the College of Arts and Sciences, and Caroline Voyles, PhD candidate and field researcher in Drexel’s Urban Sexuality Lab.

Their work examines a longstanding connection between policing and queer people of color, a population impacted by increased policing in communities experiencing homelessness, gentrification, eviction and other urban planning processes. With the help of student lab members including Nina Olney (BS ’21), Mariah Mennano (MPH ’21) and Sofia Argibay (MPH ’22), their goal is to identify and shine a spotlight on the health consequences of police oppression on this population.

By using the National Longitudinal Study for Adolescent Health, which has followed 20,000 adolescents who were in 7th to 12th grade during the 1994–1995 school year, the study will examine the relationship between sexuality, race and policing on mental and physical health through three aims:
• To identify how health outcomes relate to sexual and racial identities and determine which health outcomes are most strongly related with minority status.
• To examine how the effects of police encounters vary by sexual and racial identities and mediate the relationship between health outcomes and sexual minority and racial minority status.
• To determine a sociological theory of the effect of policing through this intersectional lens. This aim also prepares the study as possible pilot data for a future National Institutes of Health grant submission.