Beyond traditional research ethics

Ethics in research must go further than being a hurdle of compliance. Done right, it can be a partnership to guide us to better, more meaningful relationships with the people we seek to serve. This year, IDinsight has updated its ethics policy to better serve its mission of maximizing positive social impact. Our work in global development—translating data into real-world change— involves value-based decisions with significant consequences. Hence, the ethics of our actions and the dignity of the populations we affect are central to everything we do.

Traditional research ethics are largely rooted in biomedical research and the core principles of the Belmont Report to protect human subjects, and focus on academic applications. Consequently, they often fall short in considering the ethics of populations beyond research participants and more modern approaches such as those using artificial intelligence. 

Our new policy addresses these shortcomings by extending our ethical lens beyond research participants to additionally include all people impacted, our partner organizations, and project staff. Furthermore, we specifically address ethical issues arising in social science topics and from data-driven approaches.

Illustrative scenario: Introducing AI in public schools

To illustrate the inadequacy of traditional approaches to new questions arising in research ethics, consider a scenario where IDinsight partners with a government to test an AI-driven instructional tool in under-resourced schools to determine if it improves student learning outcomes.

A traditional ethical review would focus exclusively on protecting research participants: obtaining  informed consent from the teacher, parents and students and ensuring adherence to strict data collection, security and privacy protocols. 

But what about risks to other populations? Consider these questions:

• Could the tool widen educational inequalities if some students benefit while others fall further behind?
• Do all schools have adequate infrastructure to support the technology, or will the digital divide compound existing disadvantages?
• What happens to teachers whose methods are displaced or devalued by the AI system and how does this affect the broader community?
• Does the intervention create unsustainable dependency, leaving schools worse off when external support inevitably reduces or ends?

Traditional ethical review protects research participants but overlooks these systemic risks to broader populations. 

To answer these questions systematically, we developed an expanded ethics policy at IDinsight.

Key elements of the new policy 

Traditional research ethics policies often become bureaucratic compliance hurdles, are narrow in scope and can be divorced from reality. We aimed for something different: a framework deeply aligned with our values, practical for everyday use, flexible enough to span all our work, yet robust enough for rigorous ethical review. 

We do this by extending the Belmont Principles of Respect for Persons, Beneficence, and Justice to four spheres of influence:

• Project participants: Human research subjects and participants of projects.
• Impacted populations: Considering the effects on entire communities.
• Partner organizations: Ensuring equitable partnerships and sustainable impact.
• Research staff: Prioritizing their safety and well-being.

This creates a simple yet powerful framework that is intuitive to understand while grounded in comprehensive ethical foundations. Central to this approach is dignity, a core element of Respect for Persons. We put a particular emphasis on projects relying on data-driven approaches such as machine learning and artificial intelligence. 

Our goal is for every IDinsight team member to naturally use this framework when working on projects instead of only engaging with it during an ethics review.

Key challenges faced

A central challenge is balancing ethical rigour with practical constraints. Consider a project collecting data through a large-scale survey. The team wants to share de-identified aggregate results with respondents, but funding is limited and they question whether this investment is worthwhile. How should they decide?

Our framework systematically identifies risks and benefits to all affected populations throughout project implementation. The framework doesn’t eliminate difficult tradeoffs, but it helps teams navigate them with greater clarity and confidence.

Other challenges include defining key terms precisely. How broadly should we interpret “impacted populations”? How do we assess power asymmetries? Do these concepts hold across IDinsight’s diverse project portfolio, from traditional research to data science work?

Future expectations

We’re stepping into new territory and we embrace the uncertainty that comes with it. While our framework’s foundations are robust, how we assess risks, weigh benefits, and navigate real-world tradeoffs requires ongoing refinement. By stress-testing these processes on actual projects, we drive continuous improvement and ensure our approach remains relevant and practical for teams while staying deeply aligned with the values that guide all our work.

We’re not alone in grappling with these questions. We welcome dialogue with others navigating the complexities of development ethics—your feedback and collaboration will strengthen our collective commitment to rigorous, reality-grounded practice.