Well-being impact assessment of artificial intelligence-A search for causality and proposal for an open platform for well-being impact assessment of AI systems

In recent years, the well-being impact assessment approach has been applied in the area of Artificial Intelligence (AI). Existing well-being frameworks and tools provide a relevant starting point. Taking into account its multidimensional nature, well-being assessment is well suited to assess both the expected positive effects of the technology as well as unintended negative consequences. To-date the establishment of causal links mostly stems from intuitive causal models. Such approaches neglect the fact that to prove causal links between the operation of an AI system and observed effects is difficult due to the immense complexity of the socio-technical context. This article aims at providing a framework for ascertaining the attribution of effects of observed impact of AI on well-being. An elaborated approach to impact assessment potentially enabling causal inferences is demonstrated. Furthermore, a new Open Platform for Well-Being Impact Assessment of AI systems (OPIA) is introduced, which is based on a distributed community to build reproducible evidence through effective identification, refinement, iterative testing, and cross-validation of expected causal structures.