From transparency to accountability of intelligent systems: Moving beyond aspirations

Abstract A number of governmental and nongovernmental organizations have made significant efforts to encourage the development artificial intelligence in line with a series aspirational concepts such as transparency, interpretability, explainability, accountability. The difficulty at present, however, is that these exist fairly abstract level, whereas order for them tangible effects desired they need become more concrete specific. This article undertakes precisely this process concretisation, mapping how different interrelate what particular each require move from being high-level aspirations detailed enforceable requirements. We argue key concept accountability, since unless an entity can be held accountable compliance other concepts, indeed generally, those cannot do work required them. There variety taxonomies accountability literature. However, core account appears sense “answerability”; explain or give account. It ability call which provides impetus helps us understand must require.