Computer Science Curriculum: Intersectional Approaches

The Challenge

Computer science (CS) education often hones mathematical and engineering skills, while considering moral, social, and political reasoning beyond its scope. As we have seen in recent years, this can result in programs that amplify social inequities. Google Translate, for example, often defaults to the masculine pronoun when translating news articles from Spanish to English, thereby reinforcing the notion that primarily men are active intellectuals. Similarly, word embedding characterizes typical European American names as pleasant and names associated with African Americans as unpleasant—again exacerbating social biases (Zou & Schiebinger, 2018). Computer science courses that focus solely on technical programming and mathematical approaches fail to prepare students to understand how computing influences legal, governmental, economic, and cultural systems (Ko et al., 2020). Embedding intersectional analysis in core CS courses can sharpen students’ critical skills to recognize systemic injustices perpetrated by technology—and better prepare the scientific workforce for the future.

Intersectional Innovation 1: Remaking the Computing Research Ecosystem

Fostering responsible computing has become a priority in the European Union, the U.S., and elsewhere (EU, 2019). The U.S. National Academies emphasize that “with computing technologies increasingly woven into our society and infrastructure, it is vital for the computing research community to be able to address the ethical and societal challenges that can arise from the development of these technologies, from the erosion of personal privacy to the spread of false information” (National Academies, 2022). Computer researchers need to anticipate social risks from the very beginning. Failure to consider the social, political, and economic dimensions early in research can lead to harm.

Policy is one driver of technology that can catalyze structural solutions which foster social equity and environmental sustainability. This requires remaking the computing research ecosystem in a number of ways:

1) Research funders can reconfigure granting processes to support interdisciplinary work between technologists and humanists/social scientists—disciplines often siloed in different branches of funding agencies. Funding can incentivize new research partnerships (National Academies, 2022). Application evaluation needs to consider both the technical excellence and social benefits of a proposal, with special attention to gender, race, and intersectional social analysis. For example, the Human-Centered AI Institute at Stanford University requires review of research proposals by both a panel of technologists and a panel of humanists and social scientists. By putting these policies into place, research agencies help cultivate research that enhances social benefits and minimizes social harms across the whole of society.

2) Editorial boards of peer-reviewed journals and conferences can further support these efforts by requiring sophisticated critical sex, gender, race, intersectional, and social analysis when selecting papers for publication. The NeurIPS (Neural Information Processing Systems) conference, for example, conducts ethical reviews before accepting papers (Bengio et al., 2021). Journals such as Nature and The Lancet require sex and gender analysis, where relevant (Gendered Innovations, 2022).

3) Universities and research institutions can integrate knowledge of sex, gender, race, and broader intersectional social analysis into core engineering, design, and computer science curricula. Many universities host stand-alone courses on these topics in the humanities and social sciences. These are important to prepare students in those fields for collaboration, but we also need critical social analysis embedded in core courses in the natural sciences, CS, medicine, and engineering.

4) Industry. Numerous companies have promoted AI Principles similar to those articulated at the Asilomar Conference in 2017 (Future of Life Institute, 2017). It will be important to survey and audit company principles and policies for gender, race, and intersectional analysis. Industry can facilitate achieving these AI Principles by hiring employees trained to work in interdisciplinary teams that include technologists, humanists, and social scientists, and who have cultivated skills to evaluate the potential social benefits and harms of their products, services, and infrastructures.

This case study focuses on universities training, as universities train the workforce for the future. Critical intersectional social analysis must be part of core natural science and engineering courses. Efforts have been made in medicine—the Charité University Hospital in Berlin, Germany, for instance, has successfully integrated sex and gender analysis throughout all six years of medical training from early basic science to later clinical modules (Ludwig et al., 2015). (These are not materials on professional ethics; they are on sex/gender analysis in health and medicine.) However, this is a rare example, and universities must do more to prepare the scientific workforce for the future.

Intersectional Innovation 2: Emerging CS Courses

Since 2017, universities have been developing what was first called at Harvard University “Embedded EthiCS” that integrates ethical reasoning into core CS courses (Grosz et al., 2019; Garrett et al., 2020). These initiatives are also referred to as “responsible computing,” among a variety of other names. We recommend embedding intersectional sociocultural analysis that draws skills broadly from the humanities and social sciences. We highlight some emerging programs:

• ● Embedded EthiCS (Harvard University): Beginning in 2017, Professor Barbara Grosz and her CS team developed a collaboration with Professor Alison Simmons and the Philosophy department to teach CS students how to consider the ethical and social implications of their work. Embedded EthiCS introduces ethics directly into standard CS curriculum (Grosz et al., 2019). Watch their video.

  Embedded EthiCS teaches concepts such as “responsibility” applied to climate change, cloud security, performance versus correctness in system design; “rights” in relation to software verification and validation, electronic privacy and big data systems, tracking censorship; “fairness” in algorithmic fairness and recidivism prediction, discrimination and machine learning; “equality” applied to ASCII, Unicode, and the ethics of natural language representation; “discrimination” related to bias and stereotypes in word-embedding software, and more (see image). Ethics modules are available here.

• ● Embedded EthiCS (Stanford University): In 2020, Stanford initiated an Embedded EthiCS program that incorporates ethics-focused lectures, assignments, and assessments in core CS courses. The program emerged from a collaboration between the Computer Science Department, the McCoy Family Center for Ethics in Society, and the Institute for Human-Centered Artificial Intelligence (Miller, 2020).

  The goal is to develop ethics materials that relate directly to the technical course content. For example, an intro-level programming course incorporates a lecture and assignment that address the problem of bias in data sets that may lead to representational and allocation harms. Another introductory course discusses the ethics of using data for priority queues through the example of housing allocation in Los Angeles. Further, a course on algorithms discusses the potential impact of formalizing real-world problems to make them algorithmically tractable, including how foregrounding an optimization function may render other important values invisible.

• ● The Mozilla Teaching Responsible Computing Playbook outlines specific ways to foster curriculum that is successful, institutionally sustainable, and useful for students entering the workforce. Responsible computing is loosely defined as “designing computing artifacts that need to take society into consideration.” Mozilla emphasizes the need for collaboration across disciplines: “Philosophers ask about the nature of privacy and why we value it; political scientists ask how technology influences political systems; critical race theorists investigate how technology reproduces and reinforces institutional racism; psychologists consider how technology shapes the way we think and behave; disability studies analyze how disabled users ‘hack’ existing technology to make it work them. For these reasons, bringing the perspectives, concepts, and tools from other disciplines into conversation with responsible computing efforts can strongly enhance both teaching and research” (Little et al., 2021).

• ● The Impossible Project and Teaching Responsible Computing (University of Buffalo) includes a redesigned first year seminar that features coursework (lecture slides, assignments, recitation activities) specifically on critical approaches to computing. The course culminates in a competition where students pitch solutions based on a theme (e.g., antiracist computing or bias in data driven criminal risk assessment).

• ● Responsible Computing Fellows (University of Colorado Boulder): undergraduate students across campus apply for a fellowship program that involves seminar-style discussions on current events and ethical issues in technology. Students also participate in events with guest speakers and have a community of peers to discuss issues related to responsible computing, employment, coursework, and more.

• ● Responsible Problem Solving (Brown University, Haverford College, University of Utah) includes slides, assignments, instructor notes, and sample datasets for topics in data structures and algorithms related to responsible computing (Cohen et al., 2021).

• ● Embedding Ethics in CS Classes through Role Play (Georgia Tech) includes course components with written scenarios, role descriptions, supplemental readings, and teaching guides for activities related to autonomous buses and AI applications in university admissions.

• ● The Social and Ethical Responsibilities of Computing (MIT) is facilitating the development of responsible “habits of mind and action” for those who create and deploy computing technologies.

• ● Embedded Ethics (Georgetown) pairs Ethics Lab philosophers and designers with faculty across disciplines to co-design ethical discussions and run customized, creative exercises in their coursework. Students are exposed to key ethical concepts and frameworks and provided the opportunity to debate, apply, and internalize context-specific lessons in order to develop essential skills and gain a real-world understanding of their potential impacts. A partnership with the Department of Computer Science was initiated in 2019.

• ● Computing Narratives: Website with a variety of narratives that can support incorporating ethical content into computer science classes.

One should also recognize that the decision about who presents this content matters: bringing in outside experts from other disciplines can help validate the importance of social analysis. But it can also be important for this content to be taught by the CS faculty themselves to signal to students that it is important and relevant.

Intersectional Innovation 3: Inclusive Language and Visualization in Course Content

Concerted efforts have been made to make engineering and CS more appealing to traditionally underrepresented groups. A key approach has focused on breaking down stereotypes (National Academy of Engineering, 2008).

In addition to increasing the numbers of underrepresented groups, it is important to also change the content and methods of teaching and interacting within the classroom. Many concepts in computer science rely on language that reinforce historical inequities. Terms like “master” and “slaves” to describe server architecture in a distributed system (fig. 1) or using “so easy your mom can do it!” as a benchmark for user experiences are a few examples.

Figure 1: GitHub is a code hosting service for software development that provides version control. Historically, the first branch for a new code repository was named the “master” branch. However, conversations within the Git project and the Software Freedom Conservancy led to changes within GitHub to rename the default branch of the repository “main.”

Similarly, many benchmark images, such as Lena, a standard test used image in the field of computer vision since 1973, are problematic. Lena is a picture of the Swedish model Lena Forsén cropped from the centerfold of the November 1972 issue of Playboy magazine. Not only is the image sexist, but it sets “white” as the standard for image processing, which has led to many problems in facial recognition and other processes and programs (Buolamwini & Gebru, 2018). Rethinking language and visual representation serves as a relatively simple intervention to begin to challenge inequitable norms in the classroom.

Next Steps

• 1. Evaluate Embedded EthiCS and responsible computing courses. More research is needed on how to effectively evaluate courses that embed sophisticated intersectional social analysis (Horton et al., 2022).
• 2. Build embedded ethics or responsible computing into the CS major.