Scholarlink

<p>In this paper, we adopt the iconic tripartite structure of Leone’s 1966 film “The Good, The Bad, and The Ugly” to examine Artificial Intelligence as a sociotechnical phenomenon: we explore the transformative positive opportunities AI affords (the good), characterise the persistent technical and operational limitations researchers are still actively addressing (the bad), and analyse the deliberate misuses and ethical violations that arise from human choices in AI development and deployment (the ugly). Crucially, we argue that the bad and the ugly are not always distinguishable in practice, and we formalise this claim through two original contributions: the Indistinguishability Thesis, which demonstrates that bad and ugly AI outcomes can be phenomenologically identical to affected communities; and the Ethical Responsibility Gradient (ERG), a four-zone analytical framework that maps AI harms along the axes of actor agency and epistemic state to guide context-sensitive governance. Together, these contributions move the field beyond binary good/bad framings toward a graduated, sociotechnical account of AI ethics capable of supporting more targeted regulatory and accountability responses.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

The good, the bad and the ugly in artificial intelligence

  • 期刊论文

  • 发表日期：

    2026年6月30日

查看全文
  • Domingo Mery,
  • Gabriela Arriagada-Bruneau,
  • Jocelyn Dunstan

摘要

In this paper, we adopt the iconic tripartite structure of Leone’s 1966 film “The Good, The Bad, and The Ugly” to examine Artificial Intelligence as a sociotechnical phenomenon: we explore the transformative positive opportunities AI affords (the good), characterise the persistent technical and operational limitations researchers are still actively addressing (the bad), and analyse the deliberate misuses and ethical violations that arise from human choices in AI development and deployment (the ugly). Crucially, we argue that the bad and the ugly are not always distinguishable in practice, and we formalise this claim through two original contributions: the Indistinguishability Thesis, which demonstrates that bad and ugly AI outcomes can be phenomenologically identical to affected communities; and the Ethical Responsibility Gradient (ERG), a four-zone analytical framework that maps AI harms along the axes of actor agency and epistemic state to guide context-sensitive governance. Together, these contributions move the field beyond binary good/bad framings toward a graduated, sociotechnical account of AI ethics capable of supporting more targeted regulatory and accountability responses.