On a short comment published on February 3rd,2020 on Nature Machine Intelligence, Rafael A. Calvo, Dorian Peters and Stephen Cave argued for “advancing impact assessment for intelligent systems”.
According to these researchers, impact assessments have to be improved to be effective Human Impact Assessments for Technology (HIAT). In particular, these HIATs would have to be more human-centered, in order to take into account the ethical issues raised by intelligent systems (for instance, the impacts on “voting behaviours”). These HIATS would also have to be more dynamic and exhaustive, taking into account every aspect of intelligent systems, from the training to the response to an impact.
Still, the comment underlines that HIATs should rely on “existing ethical technology frameworks”, such as the European Union’s Ethics Guidelines for Trustworthy AI, and “include existing assessment methods”, such as the data protection impact assessment.
Source : Rafael A. Calvo et al. Advancing impact assessment for intelligent systems, Nature Machine Intelligence (2020). DOI: 10.1038/s42256-020-0151-z