What fairness means depends heavily on context, including the organization's objectives, the purpose of the system and its environment; automated decisions can treat particular people or groups unfairly, through biased objective functions, unbalanced datasets, human prejudice in training data or feedback, or bias in the product concept, the problem framing or when and where the system is deployed (ISO/IEC 23894 A.6). For each identified interested party the organization should analyse the foreseeable benefits and harms tied to fairness, for example: performance that is unjustifiably better or worse for a particular group (by age, gender or religion, for instance); a representation of a person that discriminates unfairly on the basis of groups associated with them; accessibility, so that people are not shut out by disability or lack of digital skills; and the effect of unwanted bias on at-risk groups.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.