Announcing Bias Bounties at Scale
In partnership with Radiant Earth and with support from Heising-Simons Foundation, we’re moving our bias bounty program onto Zindi
Co-authors: Mala Kumar, Annie Brown, Celina Lee, Jed Sundwall, Cara Franson
Thanks to the generous support of the Heising-Simons Foundation, Humane Intelligence nonprofit is excited to announce that we are partnering with Radiant Earth to move our bias bounty program onto Zindi, a global data science platform with users in more than 185 countries!
With support from Google.org, Humane Intelligence pioneered the concept of “bias bounties,” which creates structured data science challenges to identify, define, and address algorithmic bias in data. To-date, we have run four challenge sets over the past two years, most recently focusing on Accessibility in conferencing facilities. Our challenges have demonstrated both community enthusiasm and a real need to address algorithmic bias in transparent, data-driven ways, especially as generative AI solutions continue to pilot and scale across high-stakes domains, such as climate change, economic wellbeing, and gender rights.
As we detailed in a concept note earlier this year, moving the bias bounty program onto Zindi will help us:
- Lower costs and achieve faster scale by standardizing participant screening and submission scoring.
- Expand our community-driven participation, including to those without advanced degrees or certifications.
- Quantify bias metrics by creating standardized scoring frameworks and APIs.
- Release more results under open source software, content, and data licenses.
We’re excited for this new chapter in our bias bounty program!
Our Pilot Scope
The theme of our pilot bias bounty on Zindi will be geospatial and climate adaptation domains, leveraging Radiant Earth’s Source Cooperative data publishing utility. As part of the process, the three partner organizations will form a working group with climate scientists to help identify the most impactful challenge focus, based on scientific relevance and needs. These experts will also contribute to the challenge design and will help facilitate collaboration with impacted communities throughout the ML training and evaluation process. Three potential focus areas for the pilot bias bounty on Zindi are currently under consideration:
- Urban–Rural Mapping Accuracy, exploring performance disparities between urban and rural or Indigenous regions;
- Climate Adaptation and Indigenous Knowledge, examining whether model outputs reflect locally-sourced ecological and natural disaster resilience strategies; and
- Gender and Power in Geospatial Tools, investigating how patriarchal assumptions may be encoded in adaptation frameworks, embeddings, or labeling processes.
Each of these focus areas lends to a challenge structure that could include both technical and community tracks—enabling participation from ML engineers, data scientists, and members of communities most impacted by AI decisions. Challenge tracks may include qualitative audits, data annotation, metric development, and inclusive design proposals.
The project will produce public goods in the form of open source scoring frameworks, qualitative measurement APIs, and challenge design templates that make bias more visible, trackable, and addressable across AI systems. By embedding community participation into the core of AI evaluation platforms like Zindi, the bias bounty model strengthens both the social legitimacy and technical quality of AI. In doing so, this pilot will lay the groundwork for a global, replicable model of participatory accountability—one that supports both immediate system improvements and long-term shifts toward equity in AI development and deployment.
The pilot bias bounty challenge on Zindi is anticipated to launch in Q3 of 2026.
How to Get Involved
Humane Intelligence, Zindi, and Radiant Earth will be focused on internal project planning until the end of 2025. In early 2026, we will post information on our volunteer page about how to get involved in the climate working group and other aspects of the overall project. We’ll continue to post project updates on our blog throughout 2026.
About the Project Partners
Mala (Interim Executive Director, Humane Intelligence), Celina (Co-Founder and CEO of Zindi), Jed (Executive Director of Radiant Earth) at lunch during UNGA / Climate Week, Sept 2025
Humane Intelligence is a 501(c)(3) nonprofit dedicated to breaking down barriers to AI deployment for social good. We collaboratively design and run rigorous evaluations that make AI systems more accountable, responsible, and fair. Our mission is to empower people and organizations that represent or have lived experience in a given problem space, through our technical AI evaluation design, our convening events, and our red teaming web application. To date, the organization has run four bias bounty challenge sets, which were scoped across three topical areas: transparency, extremism, Indian forestry, and accessibility in conferencing facilities.
Zindi is the leading professional network and AI challenge platform for data scientists and AI builders in emerging markets. Born in Africa, the organization has rapidly expanded into a vibrant global ecosystem dedicated to making data science and artificial intelligence accessible and impactful everywhere, with more than 90,000 users in 185 countries participating. More than 60% of Zindi’s global user base is in 50+ sub-Saharan African countries and 28% are women. Zindi has awarded more than USD 1,000,000 in challenge prizes.
Radiant Earth is a technology nonprofit that increases shared understanding through community-led initiatives that make data easier to access and use. Founded in 2016, the organization works at the intersection of governance, technology, and community to enable cooperation on global challenges.
Through Source Cooperative (Source), Radiant Earth’s community-governed data publishing utility, the platform hosts over 2 petabytes of data across more than 300 data products, serving over 100 million data requests per month. Source serves as foundational infrastructure for AI development, providing simple URLs, file browsers, and programming tools that allow researchers to publish and analyze data at any scale, including weather models, satellite imagery collections, and other large archival datasets needed for AI research.
For information about the Heising-Simons Foundation, please visit: https://www.hsfoundation.org/