An Ecosystem Approach to Ethical AI and Data Use: Experimental Reflections

Ethics documents alone don't reach practitioners: The paper identifies a significant gap between the proliferation of AI ethics guidelines and their actual adoption by AI practitioners, suggesting that top-down ethical frameworks have limited impact on those building AI systems in practice.

Practitioners need practical, operational language for ethics: Rather than abstract principles, AI developers require concrete guidance tied to their actual decision-making processes. The paper proposes bridging this gap through a "shared fairness approach" that translates ethical concepts into the language and workflows practitioners use daily.

Grassroots dialogue is essential for meaningful ethics: The methodology centers on facilitated conversations with practitioners to identify when ethical challenges actually arise in their work and to help them internalize responsibility, rather than imposing ethics from external authorities.

Power imbalances undermine ethical implementation: The paper highlights how structural decision-making power and elite technical knowledge create responsibility gaps. Addressing this requires returning ethical discourse to those "at the sharp end" of implementation rather than concentrating it among leadership or ethics specialists.

AI for Social Good initiatives need practitioner input: The paper emphasizes that the growing number of AI for Social Good projects represent an emerging opportunity to develop more meaningful ethical approaches, but only if they actively involve the people actually building these systems in the ethical planning process.