Data Ethics Club: “It’s Not Exactly Meant to Be Realistic”: Student Perspectives on the Role of Ethics In Computing Group Projects#

What’s this?

This is summary of Wednesday 5th February’s Data Ethics Club discussion, where we spoke and wrote about the paper “It’s Not Exactly Meant to Be Realistic”: Student Perspectives on the Role of Ethics In Computing Group Projects by Michelle Tran and Casey Fiesler. The summary was written by Jessica Woodgate, who tried to synthesise everyone’s contributions to this document and the discussion. “We” = “someone at Data Ethics Club”. Huw Day helped with the final edit.

Article Summary#

In recent years, higher education bodies have begun incorporating ethics into computer science courses. To investigate how students view the role of ethics in their computer science education, the authors conducted focus groups with teams of students currently working on group projects. As group projects in computer science courses are designed to simulate what working in industry would be like, if there is an absence of ethical considerations in project-based courses it risks implying that one does not need to consider ethics when working in industry either.

Findings from the focus groups suggest students felt group projects to be somewhat divorced from the real world, making it difficult for them to take ethical implications seriously. Some perceived ethics as unimportant to why they were pursuing their degree – to get a job – because they felt ethics skills would either not be useful or valued in industry. To better support ethics education, participants wanted more clarity and concrete information, desiring to move away from open-ended and non-applied scenarios to learn more about current real-world tech ethics issues. Participants wanted more feedback from experts and peers, with dedicated time to think about ethics baked into the project timeline.

Recommendations from the paper for ethics education include: make projects more real; equip students with explicit resources to consider ethics; look at applied ethics in specific contexts; teach students how to think through problems and ways of thinking. Suggested best practices are to talk about ethics early, often, and with context; build ethics considerations into the timeline/structure; make ethics a priority for students; inform students why they should pretend the project is real and why ethics is important.

Discussion Summary#

How do we define ethics in the context of computing education? Is it just doing the right thing or should it involve explicitly making injustices visible?#

“Doing the right thing” is a useful catchall expression, however, it is ambiguous as ethics has many definitions. People may have expectations on one another to act ethically but in reality, assessing how to prioritise different factors is tricky. Making and learning specific rules does not equate to a sufficient ethical skillset, as contexts vary widely and the right thing to do will change depending on the setting. Instead, training people in ways of thinking and critical analysis will enable them to systematically work out the right thing to do when they encounter issues. Equipping students with the ability to conduct ethical analysis and think critically is not only important to computer science but applies to broader issues that arise from the simple fact that we exist in society. Having some more general philosophy education across the board would help students develop the ability to think through problems from different perspectives and come to reasoned decisions.

To help students ground ethics education and translate methods of thinking to the real world, starting with concrete issues is a good first step. It is important that educators provide actual insight and solutions for ethical dilemmas, instead of just stating the need for ethical tech. Looking at real examples gives an idea of the kind of issues that they may encounter, helping students relate to the different challenges that exist. In our own courses, we have been encouraged to think ethically by being shown examples of companies behaving questionably. There are many lessons we can learn from the past such as predicting income based on genes and DNA tests for IQ, despite the controversial history of IQ tests.

Effective ways of incorporating ethics education could be through ethical case study simulators or ethics emergency drills, a project by Vanessa Hanshke to simulate issues that could come up in the workplace. In the Gemstone Program, students conduct a research and design project across all four years of their degree. Students come up with the project from scratch, so they have agency and care for the project. The project is also interdisciplinary, so even if someone on the team is willing to code up whatever they think will be the quickest solution, there will likely be a social sciences or arts major there to question why they are doing it that way.

What do you think of the authors’ recommendations for designing a project based computing course with ethics learning and practise in mind?#

Incorporating ethics considerations throughout the whole pipeline makes sense to us; it seems strange to separate ethics from the activity of doing the project. Ethics essentially involves considering the impact of your actions on others, so it must be embedded early and frequently as the paper suggests. Ethics should be a key consideration in all research and ideally would coexist throughout the entire degree program, rather than only being visited in the odd lecture or project. If educators don’t really care about the topic or it feels tacked on, it will come across in the teachings.

Introducing ethics early is essential to highlight risks early. When teaching machine learning, ethics should be embedded right at the start when considering training data, as it is difficult even for the developers to understand what is happening inside black-box models. Too often, ethics is an afterthought when harm has already taken place. Projects start as theoretical and initial ideas seem unrealistic, but later end up being real projects with real consequences. For example, facial recognition was used in the estate that houses the offices of Camden Council and only brought to attention after a backlash from the general public. Another example is the medical text problem, where AI is being integrated into scientific writing. Using AI to automate writing raises concerns about misinformation and bias, which is especially consequential in the realm of health research.

When students leave education to enter jobs where they will be developing these models, it’s really important that they are equipped with skills to think through the consequences of what they are building. The study found that some students thought ethical thinking is the responsibility of other people, but developers can’t just abdicate responsibility and blame problems that arise on the data. The first thing that people should ask when commencing a new project is “should we actually do this?”.

Intertwining ethics with workflows is especially important for undergraduates who are at the start of developing their professional practices. It is understandable that students weren’t overly optimistic regarding the impact of their work, and it’s important to have a down-to-earth approach. However, through undertaking courses, students learn how to approach tasks in ways that are later useful for their software development and communication.

To fix the problems we see in society, we need to address them within the education system itself. Education should prioritise acting ethically as much as the technical aspect, making ethical thinking the only way. We need to raise people’s expectations for standard implementations and value the actions of people that behave ethically. Best practices change over time, and if students enter industry with a robust ethical skillset, it will trickle into industry norms. Bit by bit, the talented people graduating into the workforce will end up with ethics embedded in their values and (hopefully) vote with their feet e.g. by being mindful of the companies they choose to work for.

The recommendations of the paper provide insight for how the education system can improve its teaching of ethical practices. We liked that the paper took a qualitative angle to establish the recommendations, taking into account the responses of different participants. However, we would be cautious of relying too strongly on student feedback and suggestions alone, as there are whole fields dedicated to the study of education best practices. The focus group findings could be complemented by a multi/cross/joint disciplinary approach. The study would be further improved by highlighting the industry perspective, for example, by inviting industry specialists to talk about real world examples. In our own industry experience, we do already have ethics questionnaires as a standard part of a project lifecycle. These questionnaires must be approved before development can continue on an artefact. Whilst the efficacy of this might be questionable, it is a standard part of the workflow.

How can we make ethics a functional requirement of good DS/SE practises in industry?#

Bringing ethics into real-world applications takes time, and there are difficulties with translating what is learnt in theory into practice. Convincing people with power to make decisions that prioritise ethics can be challenging. Analysis paralysis can also occur with ethics, where too much time is spent discussing issues, preventing progress from being made.

Cultivating good ethical practice is supported by incorporating insights from different cultures and ensuring projects are not detached from the deployment contexts and customers. The importance of maintaining connection with context is relevant to many fields, from biology and medicine to software development, but requirements of stakeholders might be different depending on the field. In practice, embedding ethics throughout a project could be supported by integrating ethics into agile methodologies, or introducing real people into the workflow (e.g. good UX practice). The Data Hazards Project also provides useful tools for practical ways to provoke ethical thinking.

Understanding how to allocate responsibility an important ethical skill that students should be educated in. For example, when developing ethical software it is important to consider whether responsibility should fall on the engineers, project managers, or others. If people have accountability, they might be more invested in ethical practice. Being disempowered lets people off the hook. Ideally, the education students receive would help them to empower themselves in the workplace so that they feel motivated and able to bring up issues that arise. Graduates should at least be aware of the range of software roles they might enter into and issues they might encounter in these roles.

Ensuring education makes ethics part of what it means to be a “Good Engineer” supports ethics becoming a functional requirement of industry practices. However, there is a limit to the amount of good that can be done with education alone as true integration of ethics in industry requires a big cultural shift. People need to be incentivised to keep up good practices after they graduate. If ethics isn’t a part of professional expectations, it is much harder for students to engage with learning about it. The content that students learn in their courses needs to be matched by an industry appetite to apply ethics in the real world. Otherwise, people will learn the content at university and then never actually use it. For example, most students are taught how to use git, yet many don’t use what they have learnt in their jobs.

There needs to be a shift towards valuing ethics as equally important as coding and should be included in job specifications. Prioritising ethics helps bring suppressed aspects to the surface and mitigate the tendency for utopian thinking in tech that it can solve all problems. Students need to be shown how best practices are an important tool in their arsenal, helping companies avoid unethical consequences that bring negative attention.

Despite the benefits of incorporating ethics, to the company as well as individuals within it and society at large, career paths increasingly emphasise technical specialism as detached from social responsibilities. Within the workplace, high expectations on individuals impact ethical interests and ability to spend time thinking through dilemmas. This reflects an extremely fundamental and underlying ethical issue, wherein too much focus is place on lone actors to take ethic responsibility vs. the team or company as a collective. Large projects can be equally, if not more, guilty than the individuals within them. The separation of concerns further contributes to a lack of accountability.

At the end of the day, the importance of ethics in a job comes down to company culture. After the Cambridge Analytica scandal, ethics became an important angle for Facebook in recovering its reputation loss. Yet, the sincerity of these initiatives at Facebook and other big tech companies has been repeatedly questioned. The veil is now being dropped as companies draw back on ethics initiatives that were doomed to fail in response to political regime changes. The U-turns imply that these companies never truly believed in diversity bringing value but were primarily interested in reputation-washing. As a bonus, the staff who onboarded thinking these companies were a safe space for them to work in are now likely to quit, instead of being laid off. Given the troubling direction of travel for big tech companies, where “move fast and break things” has real implications for people’s actual lives, people need to be explicitly and emphatically confronted with the reality of ethical implications.

Trusting companies to hold themselves to account is an approach that we have seen repeatedly fail. To hold an industry to account, there must be appropriate regulation in place. However, regulation across jurisdictions is tricky and there will need to be some global consensus.

Attendees#

  • Huw Day, Data Scientist, University of Bristol: LinkedIn, BlueSky

  • Jessica Woodgate, PhD Student, University of Bristol

  • Euan Bennet, Lecturer, University of Glasgow, BlueSky

  • Olgierd Zagozda, Cognitive Science Student, Adam Mickiewicz University in Poznań, LinkedIn

  • Hessam Hessami, Data scientist

  • Paul Matthews, Lecturer in Data Science, UWE Bristol Mastodon

  • Robin Dasler, Data Product Manager, LinkedIn

  • Kamilla Wells, Citizen Developer, Australian Public Service, Brisbane