Software engineering for responsible AI: an empirical study and operationalised patterns

AI ethics principles and guidelines are typically high-level and do not provide concrete guidance on how to develop responsible AI systems. To address this shortcoming, we perform an empirical study involving interviews with 21 scientists and engineers to understand the practitioners' views on AI ethics principles and their implementation. Our major findings are: (1) the current practice is often a done-once-and-forget type of ethical risk assessment at a particular development step, which is not sufficient for highly uncertain and continual learning AI systems; (2) ethical requirements are either omitted or mostly stated as high-level objectives, and not specified explicitly in verifiable way as system outputs or outcomes; (3) although ethical requirements have the characteristics of cross-cutting quality and non-functional requirements amenable to architecture and design analysis, system-level architecture and design are under-explored; (4) there is a strong desire for continuously monitoring and validating AI systems post deployment for ethical requirements but current operation practices provide limited guidance. To address these findings, we suggest a preliminary list of patterns to provide operationalised guidance for developing responsible AI systems.