By Dr. Chirag Arora (c.arora@tudelft.nl) & Dr. Juan Durán (J.M.Duran@tudelft.nl)
The rapid growth of large-scale datasets and computational methods has transformed social science research. Researchers can now study collective human dynamics at a scale and resolution that was unimaginable just a decade ago. But this shift also introduces ethical challenges that traditional research frameworks were never designed to handle.
To help researchers navigate this complex landscape, colleagues at TU Delft have developed the Ethical Reflection and Consultation Tool — a collaborative service that gives researchers a structured space to explore the ethical dimensions of their work alongside knowledgeable peers.
Why existing ethical frameworks fall short
Most ethical guidelines used in social science research were developed over decades, focused on individual level, in-person studies with designed interventions. They focus heavily on direct harm to individuals, anchoring key concepts, such as the autonomous subject and informed consent. These concepts don’t map cleanly onto computational social science (CSS), where causality is often diffuse, and harms can be collective or societal rather than individual.
The Ethical Reflection and Consultation Tool fills this gap. Rather than walking researchers through procedural compliance with privacy regulations, it creates space for substantive reflection on the specific ethical dimensions of each project.
Conceptual clarity: the hidden ethical risk
One key theme the tool explores is conceptual clarity: the recognition that research concepts are rarely neutral. They carry implicit values that shape research outcomes.
A recent study on algorithmic prediction makes this concrete. A predictive model failed to flag a student’s low GPA because it relied on a single survey variable: her rating of closeness with her mother as “not very close.” That crude category completely obscured a history of severe trauma and abuse directly affecting her academic performance. No algorithm can compensate for that kind of conceptual limitation.
The tool helps researchers catch these problems early by prompting them to question the assumptions embedded in their metrics, and to ask explicitly what aspects of reality their chosen indicators might obscure or distort.
When defining fairness becomes a value choice
The tool also helps researchers grapple with fairness, a concept that sounds straightforward but is deeply contested in practice.
There is no technically correct definition: prioritising equal meaning of scores across demographic groups and prioritising equal error rates are both mathematically valid interpretations, yet mutually incompatible. The choice between them reflects specific ethical commitments and shapes whose interests the model ultimately serves.
The Ethical Reflection and Consultation Tool helps researchers examine that choice explicitly by asking what their chosen fairness metric actually optimises for, what it leaves unaddressed, and whether that aligns with the values their research is meant to uphold.
The wider consequences of data-intensive research
Computational social science research rarely affects only the people whose data it uses. The Ethical Reflection and Consultation Tool helps researchers think through how their work can affect communities that were never direct participants, and how marginalised groups face disproportionate exposure when research is repurposed or misapplied.
The risks are easiest to see in concrete cases. A study using mobile phone data to map informal employment among a vulnerable migrant population might be designed to help NGOs target aid more effectively. The same dataset, in a different legal or political context, could be used for surveillance or enforcement.
Through the tool, researchers are prompted to trace these possible trajectories: who else might be affected, how might outputs be repurposed, and what power asymmetries shape how results are likely to be used in practice? These are questions that standard ethics review processes rarely ask, and that are far easier to address before a project is underway than after.
How the process works
The tool is designed to be collaborative and low-barrier:
- Researchers complete an online form (around 30 minutes), providing context for a confidential follow-up conversation.
- A reviewer reaches out to schedule a discussion exploring the issues raised in the submission.
- The conversation acts as a sounding board, not a formal review. Final decisions about research methodology and ethical approach always remain with the researcher.
The aim is to help researchers identify potential ethical issues early, when they are still easy to address, saving time and strengthening the methodological foundations of their work.
The CSS Ethics Primer
To support the process, a companion resource — the Computational Social Science Ethics Primer — provides deeper background on common ethical issues in data-intensive research. Structured around the same themes as the reflection tool, it can be used to prepare before the consultation or as a reference during ongoing research design.
Building a culture of ethical practice
Beyond individual projects, the longer-term ambition is to build a shared vocabulary for ethical practice in CSS. If you are working on a data-intensive social science project and want a structured space to think through its ethical dimensions, we encourage you to explore the tool.
Photo by Mick Haupt on Unsplash