Introduction to Coarse Ethics: Tradeoff Between the Accuracy and Interpretability of Explainable Artificial Intelligence

Abstract

Abstract As devices powered by artificial intelligence (AI) become increasingly prevalent, the following question arises: How can these technologies improve human well-being? Transparency, defined as traceability and explainability under the European Union’s Artificial Intelligence Act, is vital to human well-being. Given the citizens’ varying levels of education and understanding, it is unrealistic to expect a single, uniform eXplainable AI framework to the public; a variety of explanations is necessary. This issue, long known as the tradeoff between AI accuracy and interpretability, remains unresolved. This chapter addresses this topic from a fresh perspective: coarse ethics. The author contends that full traceability is essential for professionals, whereas a coarse explanation of an AI system is sufficient for ordinary citizens and those with comprehension difficulties. Consequently, transparency should encompass not only a narrow concept centered on traceability and accountability but also a basic foundation that non-experts can comprehend. Under this broad definition, an AI must be explainable to the extent that it allows for reasonable informed consent. Adopting this method will alleviate the explanatory burden on developers while avoiding information overload for users and regulators, thereby promoting human well-being.

Ähnliche Arbeiten