Artificial intelligence and machine learning have made impressive strides in the last decade, with innovations that have entered the daily life of billions of people, and no signs of slowing down. Given the magnitude of its impact, the social good should not be an afterthought: market forces alone may not guarantee that these technologies benefit everyone. Instead, we believe that AI should empower those already championing humanitarian and development causes. In order to accelerate adoption of AI methods where their impact on the social good is largest, we propose to bring together non-governmental organizations working in international development and on humanitarian issues, with technical experts (academics, researchers, data scientists, engineers).

Primary objectives of this Dagstuhl Seminar are to establish partnerships and build trust, to iterate on concrete problems in a hands-on hackathon, and to demonstrate what is feasible today via case studies. Secondary objectives include scoping out new research challenges for the AI community to bite their teeth into, sharing methodological insights and publicizing efforts in the AI for Social Good space more generally. And of course, publication impact is substantially enhanced when a method has real-world impact.

We believe that the intimacy of the Dagstuhl venue is perfect for constructive communication and exchange. Planned outcomes are:

• Direct impact for NGOs by bringing state-of-the-art ML techniques to bear on their challenges, including concrete pilot showcase(s) developed in the hackathon part of the seminar.
• New research directions in machine learning that are grounded in today’s and tomorrow’s needs of NGOs (e.g. missing data, side-effects, sparse feedback, multiple competing objectives)
• New collaborations between NGOs and academics (possibly via their students) to create opportunities for long-term research that don’t end with the seminar.
• A white paper that proposes a set of specific ML method challenges that NGOs face.
• Facilitate future meetings by reflecting on the interdisciplinary process, extracting guidelines, identifying common challenges and disseminating them, e.g. in the form of a handbook .
• Visibility and acceptance of these ideas within the NGO sector and the AI community at large.

Creative Commons BY 3.0 DE

Claudia Clopath, Ruben De Winne, Rayid Ghani, and Tom Schaul