This chapter provides a meta-level self-reflection of research on AI-based social assessment with regard to its handling of normativity. It is motivated by the idea that, on the one hand, there cannot be a one-fits-all solution to ethical issues due to the high diversity of cultures and contexts. On the other hand, it is necessary to discuss the possibility of a common position based on shared fundamental norms such as non-discrimination and fairness. These efforts, reflected by the idea of establishing an Ethical Observatory, should be relevant to any research on AI-based social assessment. This chapter, therefore, explores various ethical issues related to AI in the context of social service provision, linking them to a reflective approach to social justice and how it could or should be improved by better AI. As different cultural values claim moral validity, it is necessary to scrutinise the relationship between their contextuality and universality. On this basis, discourse ethics is proposed as a model for dealing with cultural differences without abandoning universally valid standards, linking it with the role of safe spaces and stakeholder dialogues. Furthermore, it shows how the justification of moral norms can be reasonably argued. Based on this, important research ethical aspects will be considered.