The Impact of Data Augmentation and Oversampling onCultural Competence in Social Robotics

In recent years, as robots have gained popularity, concerns have emerged about their ethical and social implications, particularly the need for robots to adapt to different cultures by being culturally competent. This issue is especially pronounced in autonomous systems that use machine learning (ML), as these systems tend to mirror disparities in the data, often underrepresenting certain cultures. To address this issue, various techniques are used in ML; in this article, we apply data augmentation (DA) and oversampling (OS) to assess how these methods affect the cultural competence of ML models in two social robotics use cases.