Bias Bounty 3 marked a meaningful step in applying responsible AI to the field of environmental sustainability. In partnership with the Indian Forest Service (IFS), Government of India, this round focused on addressing bias in AI-driven reforestation recommendations—ensuring that predictive models are not only technically sound, but also fair, biophysically informed, and aligned with community priorities.

Participants explored:

• Feature importance in ecological datasets

• Predictive modeling for sustainable reforestation

• Contestability and fairness in algorithmic recommendations

This challenge emphasized the role of inclusive, transparent AI in shaping environmental policy—particularly when it impacts biodiversity and local livelihoods. Here are reflections from our Bias Bounty 3 winners:

Yashashree Garge – First Place, Thought Leadership Level

"I'm honored to receive this recognition in the Thought Leadership category. I hope this contributes meaningfully to real-world decision-making and strengthens environmental education, particularly for developers. The intersection of technology, governance, and ecological responsibility is crucial in shaping sustainable solutions. By fostering critical discussions and addressing biases in environmental decision-making, we can move toward more inclusive and effective policies.

I am extremely thankful as I have received the prize. I hope I can contribute in a meaningful way.

Thank you so much for this opportunity."

Mark Schutera – First Place, Beginner Technical Level

“Bias is not a glitch in AI – it’s part of its very nature and our techno-social task to provide guardrails.”

Mayowa Osibodu – First Place, Intermediate Technical Level

“This challenge felt very open-ended, and eventually for me it helped to not fixate on using any one approach or AI tool. Instead it helped to take a step back, try to really understand the problem involved, and be open-minded about technical components which could be pieced together to form a relevant solution."

Nagesh Mohan – Second Place, Intermediate Technical Level

"During the Bias Bounty 3 challenge, I had the opportunity to apply machine learning techniques to solve a real-world problem — identifying key factors influencing tree planting feasibility.

I applied various machine learning techniques based on the dataset, such as Decision Tree, Linear Regression, Logistic Regression, Naive Bayes, and XGBoost, to rank the importance of different factors affecting tree plantation.

Building on this, I developed a site recommendation engine that predicts site suitability for tree planting in specific regions by integrating those identified features. I also implemented techniques to identify and mitigate bias, ensuring that the model’s predictions were fair and robust across different conditions. Additionally, I focused on contestability, ensuring the model's predictions were transparent and open to challenge.

This was a valuable learning experience that enhanced my skills in analyzing raw data using machine learning models, evaluating and ranking features for tree plantation, and creating a solution to help stakeholders identify suitable planting sites. It also deepened my understanding of responsible AI practices and the importance of building fair, explainable models for real-world applications."

Bias Bounty 3 spotlighted the growing importance of integrating fairness and local context into environmental AI tools. Congratulations to all the winners—and thank you to every participant who contributed their time, ideas, and expertise.

Stay tuned for future Bias Bounties as we continue building a more inclusive, transparent AI future.