The overall rapid increase in the use of artificial intelligence technologies (AIT) is linked to various initiatives that propose AI ‘for good’. However, a holistic, in-depth understanding of these AI use cases and their implications for people, society, and the planet remains lacking, beyond the respective projects’ self-descriptions. To close this gap, we present a framework for a holistic yet manageable approach for running AI projects to assess their impact. We call our approach the Impact AI Method, as we not only want to assess the impacts of AI use cases but also want to impact AI and its use in practice and in public debate. The Impact AI Method evaluates concrete AI use cases in terms of their contribution to sustainability and public interest, reflecting these two concepts as well as established international guiding principles such as democratic values and human rights. The interview- and survey-based methodological approach combines and extends existing assessment approaches, incorporating a catalogue of 25 assessment criteria to evaluate the results. The Impact AI Method is intended as a reusable blueprint to inform public debate about AI ‘for good’ claims and to support the creation of evidence and transparency for AI systems that purport to contribute to just and sustainable development.
