Scholarlink

<p>The growing environmental impact of machine learning (ML) research calls for a systematic assessment of the community’s awareness of sustainability issues. In this study, we analyze over 1,100 papers published between 2021 and 2024 in the leading ML conferences (NeurIPS, ICML, ICLR, AAAI, IJCAI, ECML-PKDD) to assess how environmental sustainability is addressed. Using a structured classification framework, we examine key aspects such as algorithmic choices, hardware configurations, cloud usage, optimization techniques, carbon emissions reporting, and the presence of proposed mitigation strategies. Our findings reveal a limited level of awareness: only a small fraction of papers explicitly mention environmental concerns, and emissions reporting is almost entirely absent. We also identify significant gaps in the documentation of computing resources and sustainability strategies. This study provides a critical overview of current research practices and promotes the adoption of sustainable guidelines in ML, contributing to a more environmentally responsible AI research ecosystem.</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

Assessing awareness of environmental sustainability in machine learning research

  • 期刊论文

  • 发表日期：

    2025年6月10日

查看全文
  • Elio Masciari,
  • Enea Vincenzo Napolitano

摘要

The growing environmental impact of machine learning (ML) research calls for a systematic assessment of the community’s awareness of sustainability issues. In this study, we analyze over 1,100 papers published between 2021 and 2024 in the leading ML conferences (NeurIPS, ICML, ICLR, AAAI, IJCAI, ECML-PKDD) to assess how environmental sustainability is addressed. Using a structured classification framework, we examine key aspects such as algorithmic choices, hardware configurations, cloud usage, optimization techniques, carbon emissions reporting, and the presence of proposed mitigation strategies. Our findings reveal a limited level of awareness: only a small fraction of papers explicitly mention environmental concerns, and emissions reporting is almost entirely absent. We also identify significant gaps in the documentation of computing resources and sustainability strategies. This study provides a critical overview of current research practices and promotes the adoption of sustainable guidelines in ML, contributing to a more environmentally responsible AI research ecosystem.