As social scientists who work with digital data increasingly set aside informed consent, can designers bring fundamental ethical values back into research?

That’s the question that Nathan Matias and I started asking in 2017 when designing Bartleby, new software for research ethics. Four years later, we just published a peer reviewed article about Bartleby and the ideas behind it in the journal Social Media + Society.

In 2014, the public was outraged to learn about a study that altered the contents of hundreds of thousands of Facebook news feeds (Kramer, Guillory, & Hancock, 2014). Because the Facebook Emotion Contagion study lacked informed consent, people who were included in the research data likely never found out whether or not they were in the study. Facebook’s study is just one of many examples of data collection that happens online without people’s knowledge.

Scholars suggested that the ethics of this and similar studies could have been improved with minimal effort by debriefing participants. Debriefing involves notifying people of their participation and offering a chance to opt out (Grimmelmann, 2015). Yet surveys of research practices in social computing have found that many academics avoid ever consenting or informing participants because they believe it to be impractical (Vitak, Shilton, & Ashktorab, 2016). We don’t think inconvenience is a persuasive reason to take ethical shortcuts, so we set out to do better.

inconvenience isn’t a persuasive reason to take ethical shortcuts

To advance research ethics procedures that protect individual autonomy, prevent abuses of power, and promote public trust, we designed Bartleby: a system that delivers research debriefing for large-scale online studies. Using Bartleby, researchers can automatically send each of their study participants a message directing them to a website where they can learn about their involvement in research, view what data researchers collected about them, and give feedback. Most importantly, participants can use the website to opt out and request to delete their data. The system is named after the titular character in Herman Melville’s short story Bartleby, the Scrivener. Over the course of the story, Bartleby opts out of completing various requests. Instead, he states simply that he “would prefer not to.”

Why design software for research ethics?

Designing software systems to support research ethics is important for putting ethics into practice. As scholars, we also create software to advance knowledge. Design is a way to develop ethics theories by putting theory in conversation with practices that matter to people’s lives.

design is a way to develop ethics theories by putting theory in conversation with practices that matter to people’s lives

When ethics systems are put into practice, they can enable participant voice and power. Debriefing provides participants with an opportunity to opt out that would otherwise not have existed. Debriefing can also make researchers aware of contrasting values held by different participants, informing how researchers think and act on their normative values and uses of power. Because Bartleby is available as open source software, it lowers the technical burden required for researchers to implement ethics procedures. Grimmelman argues that “as it becomes easier to do more for participants, researchers should, because there is less and less reason not to” (Grimmelmann, 2015).

Bartleby also serves as a prompt for critical reflection on how to evaluate the ethics of research. To structure researchers’ thinking about how to evaluate the ethics of a study, our paper discusses two kinds of ethical theories drawn from feminist and political philosophy: procedural and substantive theories. Procedural theories are concerned with the abilities and limitations of scalable, repeatable procedures to protect individual autonomy. Substantive theories are concerned with the values upheld by the research, and the use of power in deciding those values. As we discovered in our deployments of Bartleby, using software to scale research ethics for large studies surfaces both procedural and substantive ethical considerations.

Procedural and substantive theories in research ethics

Procedural theories focus on the details of procedures—standardized, repeatable steps that can be automated by a software system or a bureaucracy—that protect rights, such as a right to individual autonomy. Even in the absence of institutional requirements, online researchers and their participants can benefit from improvements in the design and use of ethics procedures. For example, many offered procedural criticisms of the Facebook Emotion Contagion study, arguing that it should have included consent or debriefing.

As new research areas are developed, advances in procedural ethics are required to account for new situations that arise. For example, university IRB staff have struggled with the fact that the “Common Rule … does not provide appropriate guidance for the realities of research with online data” (Vitak, Proferes, Shilton, & Ashktorab, 2017). Researchers in political science and philosophy have studied the moral significance of ethics procedures other than prior informed consent—such as proxy consent (Humphreys, 2015) and hypothetical consent (Enoch, 2017)—due to their necessity in practical situations that do not satisfy the assumed ideal conditions for informed consent.

Researchers applying substantive ethical theories would ask how participants exercise voice and power in decisions about the design, implementation, and uses of research

Researchers apply substantive theories to research ethics when they have value-driven conversations about a study’s content, design, and other ethically-relevant issues—regardless of any procedures employed. For example, a substantive ethics conversation on the Facebook study would cover topics including mental health risks, collective risks associated with large-scale attempts at social influence, and the nature of people’s relationship with Facebook.

These substantive conversations often depend on the content and context of a specific study. These conversations are also affected by potentially different views about normative concepts (such as the common good) held by participants and researchers. As conflicts are ignored or negotiated, the power structures and power imbalances that enable those moves are also a concern of substantive ethics. Researchers applying substantive theories would ask how participants are able to exercise voice and power (if at all) in normative discussions and decisions about the design, implementation, and uses of research.

Just as researchers work to weave together individual autonomy and the common good, we should also see procedural and substantive theories as complementary. Rather than supplanting each other, these theories provide resources for combining equally-important considerations inherent in research ethics.

What can researchers learn from this work?

As frequent scandals involving university and corporate researchers continue to demonstrate, doing the bare minimum to comply with research ethics regulations is not enough.

Academic institutions are currently grappling with how to make structural changes that promote deep researcher engagement with research ethics questions. IRBs are thinking about how to take procedures designed in the context of 20th century biomedical experiments and update them for online social and behavioral research (Metcalf & Crawford, 2016). Conferences and funders are experimenting with ethics review processes that are integrated into peer review or funding decisions (NeurIPS 2021, Bernstein et al., 2021).

For these efforts to be successful, researchers need clear frameworks for thinking about research ethics beyond mere compliance. Procedural and substantive theories of ethics offer one such framework that can be used as a tool for critical reflection. Researchers need to ask procedural questions about whether processes are in place to protect participants, even if IRBs don’t request them. They must also ask substantive questions about whether the values and expectations of participants align with their own, and how power and positionality play a role in how differences are resolved. Weaving together these two kinds of theories as equally-important considerations can help researchers more clearly conduct ethical reflection.