Scholarlink

<p>Artificial intelligence (AI) is often framed as the pursuit of precision, efficiency, and rationality. Yet human intelligence is marked by imperfection, errors, unpredictability, and emotional nuance. While machine intelligence often strives for perfection, this paper argues that its flaws may be what make it most human-like. Using a philosophical–analytic and case-based approach grounded in debates on rationality, explainability, and AI trust (Simon in Models of man: social and rational. Wiley, New York, 1957; Mitchell in Nature 574:S50–S52, 2019. <a href="https://doi.org/10.1038/d41586-019-03013-5">https://doi.org/10.1038/d41586-019-03013-5</a>; Russell in Human compatible: artificial intelligence and the problem of control. Viking, New York, 2019), I argue that imperfection should be reinterpreted as a design principle rather than a flaw. Specifically, the paper (1) defines imperfection as a structural property of intelligent systems, (2) develops the <i>paradox of imperfection</i> framework, linking complexity with emergent error, and (3) proposes governance principles that operationalize imperfection as a resource for ethical and adaptive design. Methodologically, the paper combines conceptual analysis with illustrative cases from predictive policing, healthcare, and autonomous systems. Together, these elements develop a unified conceptual vocabulary and a normative framework for imperfection-aware AI governance that bridges theory with sociotechnical practice. Reframing imperfection as a constitutive element of intelligence opens new pathways for ethical and accountable AI. Rather than minimizing error at all costs, future AI design should balance precision with human-like unpredictability, enabling systems that are more adaptive, trustworthy, and socially embedded.</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

Imperfection as a constitutive property of artificial intelligence

  • 期刊论文

  • 发表日期：

    2026年3月9日

查看全文
  • Joshua Sanctus,
  • Shubhang Varda

摘要

Artificial intelligence (AI) is often framed as the pursuit of precision, efficiency, and rationality. Yet human intelligence is marked by imperfection, errors, unpredictability, and emotional nuance. While machine intelligence often strives for perfection, this paper argues that its flaws may be what make it most human-like. Using a philosophical–analytic and case-based approach grounded in debates on rationality, explainability, and AI trust (Simon in Models of man: social and rational. Wiley, New York, 1957; Mitchell in Nature 574:S50–S52, 2019. https://doi.org/10.1038/d41586-019-03013-5; Russell in Human compatible: artificial intelligence and the problem of control. Viking, New York, 2019), I argue that imperfection should be reinterpreted as a design principle rather than a flaw. Specifically, the paper (1) defines imperfection as a structural property of intelligent systems, (2) develops the paradox of imperfection framework, linking complexity with emergent error, and (3) proposes governance principles that operationalize imperfection as a resource for ethical and adaptive design. Methodologically, the paper combines conceptual analysis with illustrative cases from predictive policing, healthcare, and autonomous systems. Together, these elements develop a unified conceptual vocabulary and a normative framework for imperfection-aware AI governance that bridges theory with sociotechnical practice. Reframing imperfection as a constitutive element of intelligence opens new pathways for ethical and accountable AI. Rather than minimizing error at all costs, future AI design should balance precision with human-like unpredictability, enabling systems that are more adaptive, trustworthy, and socially embedded.