Responsible data stewardship Case Studies

Sasha Morinière, Joe Massey, Jane Crowe, Elena Simperl, Josh D’Addario2024 · DOI

TL;DR

The Open Data Institute (ODI) presents a framework and eight global case studies to define and demonstrate responsible data stewardship. By moving beyond mere regulatory compliance, the research shows how an iterative, systemic approach to data collection and sharing can unlock the potential of sensitive information to address structural inequalities and provide significant public benefit.

What Problem Does This Paper Address?

In the age of AI, data is the "lifeblood" of global progress, yet its potential is often stifled by two extremes: data hoarding (refusal to share) and data fearing (avoidance of data due to perceived risks). While many organizations steward data effectively, there has been a lack of a systematic, evidence-based mechanism to define what "responsible" stewardship actually looks like in practice.

The ODI seeks to bridge the gap between theoretical data ethics and real-world application. They address the challenge of how organizations can handle sensitive or previously "closed" data in a way that mitigates harm while actively redressing societal imbalances, such as gender gaps in healthcare or the marginalization of Indigenous knowledge.

Methods

The research was conducted between September 2023 and January 2024 using a multi-pronged qualitative approach:

  • Literature Review: A thorough sweep of academic and "grey" literature (reports, white papers) on responsible data practices.
  • Stakeholder Engagement: Two workshops with data governance experts and 18 in-depth interviews with practitioners across various sectors and geographies.
  • Case Study Selection: From an initial list of 40+ organizations, the team selected eight based on geographical balance (over 40% from the Global South), sectoral diversity, and their ability to address at least two of the ODI’s five dimensions of responsible stewardship.
  • The Framework: The authors evaluate stewardship through five lenses:
  • Iterative: Constant evolution based on changing contexts.
  • Systemic: Understanding impacts beyond the organization's walls.
  • Public Benefit: Prioritizing the common good over organizational gain.
  • Reduce Harm: Going beyond legal requirements for privacy and security.
  • Redressing Structural Inequality: Actively challenging power imbalances.

Datasets

As this is a qualitative research report, it does not introduce a new computational benchmark. Instead, it "stewards" a dataset of eight diverse case studies that serve as a practical evidence base:

  1. Terrastories (Suriname): Indigenous oral history mapping.
  2. Women in Data (UK): Linking retail and clinical data for women's health.
  3. ABALOBI (South Africa): Community-led data for small-scale fishers.
  4. Sightsavers (Global): Disaggregated data on disability prevalence.
  5. Masakhane (South Africa): Open-source NLP for African languages.
  6. Wikirate (Germany): Open data for corporate accountability.
  7. Karya (India): Ethical data labeling for underserved languages.
  8. Magic Box (UNICEF): Real-time mobility data for humanitarian response.

Key Insights and Ideas

  • Stewardship as a Process, Not a State: Responsibility is defined as an iterative and negotiated process. It is not a checkbox to be ticked but a commitment to constant reflection and adaptation as social contexts change.
  • Ownership vs. Agency: Several case studies (ABALOBI, Karya) highlight a shift from data "extraction" to data "ownership" by the subjects. When fishers or rural workers own their data, it transforms from a liability into an economic and political asset.
  • The Power of Disaggregation: The Sightsavers case study reveals that "standard" data collection often misses the mark. By using the "Washington Group questions," they found a 15% disability prevalence compared to the 1% typically self-reported, proving that how you ask for data determines who is visible to policy.
  • Redressing Linguistic Inequality: Masakhane and Karya demonstrate that data stewardship can be a tool for "digital decolonization." By creating high-quality datasets for "low-resourced" languages, they allow marginalized communities to participate in the AI-driven economy on their own terms.

Key Results

The primary result is the validation of the ODI’s Responsible Data Stewardship Framework.

  • Qualitative Success: The case studies prove that sensitive data (like retail habits or Indigenous locations) can be shared safely if granular control mechanisms are in place. For example, Terrastories allows communities to set differentiated access for "restricted" cultural content.
  • Impact on Policy: ABALOBI’s data stewardship was officially endorsed as the catch management system for South Africa’s Small-scale Fisheries Policy, showing that responsible community data can directly influence national governance.
  • Economic Equity: Karya pays its contributors 20 times the Indian minimum wage and allows them to profit from the repeated sale of datasets they helped create, providing a blueprint for an "ethical alternative" to extractive data labeling.

Limitations and Future Work

The authors acknowledge several constraints:

  • Scope of Research: The study relied solely on desk research and public information; it did not involve internal audits of the organizations' private practices.
  • Language Bias: The research was confined to resources available in English, potentially omitting innovative stewardship models from non-English speaking regions.
  • Organizational Holism: Acknowledging a "responsible practice" within an organization does not imply the entire organization is responsible in all its actions.
  • Future Directions: The ODI aims to expand this research to capture a more comprehensive global landscape and continue developing tools that help organizations move from theoretical ethics to practical, systemic data stewardship.

Summary updated

Abstract

Magic Box Limitations MethodologyAccess to data is critical to tackling some of society's biggest challenges, such as the climate crisis and health inequalities.Once a by-product of industrial, commercial, consumer and other activities, data has become the lifeblood of the way the world works in the 21st century.Nowhere is this clearer than in the development of Artificial Intelligence (AI) -as without data, there is no AI.Data stewardship, which we define as the 'collection, maintenance and sharing of data', is a crucial part of enabling data to be used effectively.The concept of data stewardship is a response to 'data hoarding' and 'data fearing' scenarios, where data is not shared and thus the benefits of sharing data are not realised.At the ODI, we've explored the topic of data stewardship through our data institutions programme.Data institutions are 'organisations that are stewarding data on behalf of others towards public, educational or charitable aims'.We've mapped out the landscape of data institutions, considered how they can be financially sustainable, and measured their impact.Over the last year, the ODI has been researching how organisations can steward data responsibly.We have worked with lots of data institutions that have demonstrated how to steward data effectively, however we have not had a systematic and reliable mechanism to collect evidence that identifies what constitutes responsible data stewardship.We researched what responsibility with data means to different actors through the data ecosystem, and developed a definition of responsible data stewardship to use in our work to help others design and practise responsible data collection, use and sharing.This research involved a thorough literature review of academic, grey literature and practical resources on responsible data, alongside two workshops with a stakeholder group of experts from the field of data governance and 18 interviews with experts and practitioners working with data from across a variety of sectors and geographies.
BibTeX
@misc{moriniere2024responsible,
  author = {Sasha Morinière and Joe Massey and Jane Crowe and Elena Simperl and Josh D’Addario},
  title = {{Responsible data stewardship Case Studies}},
  year = {2024},
  doi = {10.61557/jyyr8230},
}
Figure