Artificial Intelligence Explainability

Abstract

Artificial intelligence has and continues to radically transform how we live our daily lives. However, the radicalization of modern life by AI is not solely marked by convenience and progress. The proliferation of algorithmic decision-making systems has raised concerns about bias, discrimination, and the reinforcement of existing societal inequalities. These concerns are currently and rightfully at the forefront of efforts to use AI to facilitate human performance, training, assessment, and selection. As a result, AI systems must retain a human user-centered design approach so that their output and process can be monitored, evaluated, and revised to ensure accurate, high-quality decision-making. This chapter describes explainable AI (XAI) designed via this approach; addresses the major issues currently facing its design and implementation for human performance assessment; discusses the ethics associated with these systems’ current and future use; provides recommendations for ethical design; and identifies some of the key challenges that need to be overcome in order to improve the design of human-centered AI systems intended for effective human performance assessment, training, and selection.

Ähnliche Arbeiten