Making AI systems that are transparent and explainable is an ongoing challenge. Contemporary machine learning models are built in a way that does not easily lead to explanations. Whether such “black box” approaches can be avoided in important applications is unclear and subject to ongoing research. There is a particular push towards having a right to an explanation for decisions made by governments that affect individuals. Some legislation in this area has been passed in various countries. 

How explainability affects interpersonal relationships that are impacted by AI-based decisions is not yet widely discussed. The connection between acting for reasons and morality was developed in great depth by the 18th-century philosopher Immanuel Kant, whose view on this is spelled out in accessible contemporary terms here. A more advanced survey of the computer science side of explainable AI with many further links can be found here. 

Next Section »