為何應該讓人工智能強化倫理衝突的緊急決策?

Abstract

針對緊急且涉及倫理兩難或衝突而人們難以做出良好決策的事項上，本文主張應該設計人工智能(artificial intelligence，簡稱AI)來改善甚至取代人們做倫理決策。這種AI倫理衝突決策系統，表面上是取代人類做出決策，但實質上是人類借助AI做出較佳的倫理決策，這種決策是從資料驅動到AI驅動。本文透過檢討那些反對以AI做倫理決策的方法論與本體論等兩類批評，並從檢討中指出這兩類批評，有哪部分值得AI開發者警惕，有哪部分則是對於這類倫理決策系統的誤解。本文並進一步指出，這種借助AI來改善涉及公共事務的緊急倫理衝突決策，若能從AI系統設計到部署的每個階段，皆納入決策利益相關人員甚至公眾的參與，便是辛諾特—阿姆斯壯與斯科堡於2021年所主張的「人工改良民主」。

This article argues that artificial intelligence (AI) should be designed to improve or even replace human decision-making in urgent situations involving ethical dilemmas or conflicts, which are well-known to be difficult for people to make good decisions. While the AI ethical conflict decision-making system may seem to supplant human decision-making, in reality, humans utilize AI to make better ethical decisions, transitioning from data- driven to AI-driven decision-making. By examining two types of criticisms—methodology and ontology—against using AI in ethical decision-making, this article points out which parts AI developers should be aware of and which are misunderstandings of these systems. Furthermore, the article suggests that involving decision-makers and even the public in every stage of AI system design and deployment can be viewed as an “Artificial Improved Democracy” proposed by Sinnott-Armstrong and Skorburg in 2021, which could enhance the development of AI ethical decision-making systems involving public affairs.