Computationally-intensive research (CIR) takes place on a wide variety of topics including AI. Its environmental impact is potentially significant yet it does not always fall clearly within the scope of organisational ethics review policy on its own merits. Many academic institutions have ethics oversight bodies (e.g. Research Ethics Committees or Institutional Review Boards) that occupy a potentially powerful position to encourage recognition of these issues and seek reflexive practice in researchers. However, policies are often poorly-defined in respect of environmental issues and thus research is not reviewed, reviewers have little guidance for legitimate critique, and researchers are not challenged to consider planetary limits on computing resources and the interaction of these with their research. This paper aims to address these problems by proposing scoping criteria for institutional ethics policy to bring CIR within the scope of ethics review on its own merits, framing evidential criteria for reviewers to apply in ethics review, and presenting a method by which CIR researchers can reflect on their proposed research in relation to environmental factors, and assess its potential value in the light of planetary limits.

翻译：计算密集型研究（CIR）广泛涵盖包括人工智能在内的多元主题。其环境影响力虽不可小觑，却常因自身特性而未能明确纳入机构伦理审查政策的适用范围。众多学术机构设有伦理监督机构（如研究伦理委员会或机构审查委员会），这些机构在推动学界认知此类问题、引导研究者进行反思性实践方面具有潜在关键作用。然而，当前政策在环境议题方面往往缺乏明确定义，导致相关研究未受审查、审查人员缺乏合理评判依据、研究者亦无需对计算资源的行星边界及其与研究的交互作用进行反思。本文旨在通过提出机构伦理政策的界定标准（将计算密集型研究基于自身特性纳入伦理审查范围）、构建审查人员可应用的证据评估框架，以及为计算密集型研究者提供对拟议项目进行环境因素反思、依据行星边界评估其潜在价值的方法论，系统解决上述问题。