Culture-Competent Machine Learning in Social Robotics

Nowadays, Machine Learning (ML) is largely applied to Intelligent Machines, empowering automatization. Thanks to ML, researchers were able to give birth to the field of Social Robotics (SR), namely the field of Robotics which designs robots capable of interacting socially with humans. Among the SR’s objectives, researchers try to embed cultural competence in robots. Since, in order to design culture-competent robots, it is also necessary to design culturally competent ML algorithms, this paper proposes a metric for measuring cultural competence inside ML algorithms and a mitigation strategy for embedding this property inside classical ML algorithms. The metric and the mitigation strategy are evaluated on an SR use case, which dataset has been ad hoc collected.