In AI, bias is the capacity to treat different cases differently and is fundamental to learning; in the social sense bias means unfair differential treatment, so the standard uses unfairness for unjustified differences that favour some groups, which can produce favouritism or discrimination. Unwanted bias enters unintentionally from interrelated sources: human cognitive bias affecting decisions on data collection, processing, design and training; data bias, the major source, in training data; and engineering decisions. Minimising it is hard but detecting and treating it is possible, with ISO/IEC TR 24027:2021 as the reference.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.