这是用户在 2025-8-4 16:47 为 https://app.immersivetranslate.com/pdf-pro/13521951-cdd4-4b12-8faa-067ff49215d0/ 保存的双语快照页面，由 沉浸式翻译 提供双语支持。了解如何保存？

AI and the governance of sustainable development. An idea analysis of the European Union, the United Nations, and the World Economic Forum
人工智能与可持续发展的治理。欧盟、联合国和世界经济论坛的思想分析

Marie Francisco a , b , ∗ , 1 a , b , ∗ , 1 ^(a,b,^(**),1){ }^{\mathrm{a}, \mathrm{b},{ }^{*}, 1}, Björn-Ola Linnér a , b a , b ^(a,b){ }^{\mathrm{a}, \mathrm{b}}
玛丽·弗朗西斯科 a , b , ∗ , 1 a , b , ∗ , 1 ^(a,b,^(**),1){ }^{\mathrm{a}, \mathrm{b},{ }^{*}, 1} ， 比约恩-奥拉·林内尔 a , b a , b ^(a,b){ }^{\mathrm{a}, \mathrm{b}} a a  ^("a "){ }^{\text {a }} Department of Thematic Studies, Environmental Change, Linköping University, Sweden
瑞典林雪平大学环境变化专题研究 a a  ^("a "){ }^{\text {a }} 系 b b ^(b){ }^{\mathrm{b}} Center for Climate Science and Policy Research, Linköping University, Sweden
瑞典林雪平大学气候科学与政策研究 b b ^(b){ }^{\mathrm{b}} 中心

A R T I C L E I N F O

Keywords:  关键字：

Idea analysis  想法分析
Artificial intelligence  人工智能
Sustainable development  可持续发展
Global environmental governance
全球环境治理
Discourse  话语

Abstrat  摘要

This paper presents an idea analysis of AI in the policy documents and reports of the United Nations, the European Union, and the World Economic Forum. The three organisations expect AI to contribute to sustainability and a prosperous future with better data analysis, greater amounts of quantitative knowledge, and by making economic and social activities less wasteful and more energy efficient. Several challenges are also named: ethics, human rights, cybersecurity, access to reliable data, transparency, and the digital gap. The solutions presented are multi-stakeholder collaboration, cohesive but flexible governance frameworks, but also taking the lead to push for ethical and value-based AI and making sure AI is sustainable. Ideas about AI appear to stem from discourses of ecological modernisation and green governmentality. This framing turns political and structural challenges into technical issues to be solved with more data, greater collaboration, and technical progress. The similarities in ideas between the EU, the UN, and the World Economic Forum also suggest that ideas about AI and sustainable development have reached discourse institutionalisation. Ideas about AI are therefore likely to reinforce already existing institutional and discursive settings.
本文对联合国、欧盟和世界经济论坛的政策文件和报告中人工智能进行了思想分析。这三个组织期望人工智能通过更好的数据分析、更多的定量知识以及减少浪费和提高能源效率，为可持续发展和繁荣的未来做出贡献。还提到了几个挑战：道德、人权、网络安全、获得可靠数据、透明度和数字鸿沟。提出的解决方案是多利益相关者协作、有凝聚力但灵活的治理框架，但也要带头推动道德和基于价值的人工智能，并确保人工智能是可持续的。关于人工智能的想法似乎源于生态现代化和绿色政府的话语。这种框架将政治和结构性挑战转化为技术问题，需要通过更多数据、更大的协作和技术进步来解决。欧盟、联合国和世界经济论坛之间思想上的相似性也表明，关于人工智能和可持续发展的思想已经达到话语制度化。因此，关于人工智能的想法可能会强化已经存在的制度和话语环境。

1. Introduction  1. 简介

Interest in how artificial intelligence (AI) impacts sustainability pathways and international relations has grown over the years (Briscoe and Fairbanks, 2020; Shaw, 2017; Vinuesa et al., 2020). AI is an umbrella term whose definitions are manifold (Crawford, 2021) and often includes technologies such as machine learning, neural networks, and all types of autonomous systems. AI is sometimes referred to as a general-purpose technology (Chief Executive Board for Coordination, 2020; European Commission, 2021a; Smart Africa, 2021), as its use is increasingly pervasive in a vast array of fields. This breadth contributes to a vagueness that is often reflected in policies and governance settings (Ulnicane et al., 2022). In this paper, AI “refers to a class of computer programs [and related technologies, our addition] designed to solve problems requiring inferential reasoning, decision-making based on incomplete or uncertain information, classification, optimization, and perception” (Bathaee, 2018, p. 898).
多年来，人们对人工智能 （AI） 如何影响可持续发展途径和国际关系的兴趣与日俱增（Briscoe 和 Fairbanks，2020 年;Shaw，2017 年;Vinuesa 等人，2020 年）。人工智能是一个总称，其定义是多方面的（Crawford，2021），通常包括机器学习、神经网络和所有类型的自主系统等技术。人工智能有时被称为一种通用技术（首席执行官协调委员会，2020 年;欧盟委员会，2021a;Smart Africa，2021 年），因为它的使用在广泛的领域中越来越普遍。这种广度导致了政策和治理环境中经常反映的模糊性（Ulnicane 等人，2022 年）。在本文中，人工智能“指的是一类计算机程序[以及相关技术，我们的补充]，旨在解决需要推理推理、基于不完整或不确定信息的决策、分类、优化和感知的问题”（Bathaee，2018 年，第 898 页）。
The political impact of AI on sustainability varies depending on its implementation and context thereof. AI can be an asset in safeguarding biodiversity and fostering sustainable development (Goralski and Tan,
人工智能对可持续性的政治影响因其实施和背景而异。人工智能可以成为保护生物多样性和促进可持续发展的资产（Goralski 和 Tan，
2020). Alternatively, it can reinforce behaviours behind current environmental crises (Dauvergne, 2020a). Countries like the United States, China and Russia want to leverage AI to shift the balance of power in their favour (Shaw, 2017; Dauvergne, 2021; Lee, 2018; Thornton and Miron, 2020), whereas other international actors could make sense of AI in different ways, depending on their interests, goals, and contexts.
2020）。或者，它可以强化当前环境危机背后的行为（Dauvergne，2020a）。美国、中国和俄罗斯等国家希望利用人工智能将力量平衡转向对他们有利的方向（Shaw，2017 年;Dauvergne，2021 年;Lee，2018 年;Thornton 和 Miron，2020 年），而其他国际行为体可以根据他们的兴趣、目标和背景以不同的方式理解人工智能。
The understanding of policy problems is always, at least in part, socially constructed by employing discursive tools such as storylines (Hajer, 1997) and ideas (Blyth, 2002; Gofas and Hay, 2010). This begs the question of what is meant when decision-makers speak of AI for sustainability.
对政策问题的理解总是至少部分地通过使用故事情节（Hajer，1997）和思想（Blyth，2002;Gofas 和 Hay，2010 年）。这就引出了一个问题，即当决策者谈论人工智能促进可持续发展时，这意味着什么。
Examining ideas about AI for sustainability can deepen our understanding of political and other discursive underpinnings of AI strategies and policies. This paper offers a qualitative idea analysis of policy documents and reports that directly or indirectly link AI and sustainability. It analyses ideas about AI in global environmental governance, using the European Union (EU), the United Nations (UN), and the World Economic Forum as case studies. We ask: how is “AI” framed in the selected material, and how is it articulated in relation to broader environmental and sustainability discourses?
研究有关人工智能可持续发展的想法可以加深我们对人工智能战略和政策的政治和其他话语基础的理解。本文对直接或间接将人工智能与可持续发展联系起来的政策文件和报告进行了定性的想法分析。它以欧盟 （EU）、联合国 （UN） 和世界经济论坛为案例研究，分析了全球环境治理中人工智能的想法。我们问：“人工智能”在所选材料中是如何构建的，以及它如何与更广泛的环境和可持续发展话语联系起来？
We began our study with the hypothesis that idea frames about AI and sustainability either spur change within discourses and institutions or, conversely, reinforce already existing discourses. We understand sustainability as a “people-centred and conservation-based concept that implies the development of the standard of human life by respecting nature’s capacity to afford life-support facilities and resources” (Rout et al., 2020, p. 4). Sustainable development is a strand of sustainability concerned more specifically with promoting economic growth within the boundaries of Earth’s natural systems (Portney, 2015). Our focus is on how ideas shed light on the role of AI in social and environmental wellbeing.
我们开始研究的假设是，关于人工智能和可持续性的思想框架要么刺激话语和机构内部的变化，要么相反，强化已经存在的话语。我们将可持续发展理解为“以人为本、以保护为基础的概念，意味着通过尊重自然提供生命支持设施和资源的能力来发展人类生活水平”（Rout 等人，2020 年，第 4 页）。可持续发展是可持续发展的一个分支，更具体地涉及促进地球自然系统范围内的经济增长（Portney，2015）。我们的重点是想法如何阐明人工智能在社会和环境福祉中的作用。

2. Theoretical and methodological background
2. 理论和方法论背景

2.1. Why should we approach AI from an idea perspective?
2.1. 为什么我们应该从想法的角度来处理人工智能？

Science and technology scholars have long studied the political impacts of technologies (Eubanks, 2011; Feenberg, 2002; Winner, 1980). Despite some notable exceptions (Ulnicane et al., 2022; Thornton and Miron, 2020; Cath et al., 2017; Ulnicane, 2022), the literature on international relations and sustainable development has not engaged with AI in terms of discourses or political ideas, even when political effects on democracy, equity and ethics are recognised (see for example Shaw, 2017; Lee, 2018; Nour, 2019).
科学和技术学者长期以来一直在研究技术的政治影响（Eubanks，2011 年;Feenberg，2002 年;Winner，1980 年）。尽管有一些值得注意的例外（Ulnicane 等人，2022 年;Thornton 和 Miron，2020 年;Cath 等人，2017 年;Ulnicane，2022 年），有关国际关系和可持续发展的文献尚未在话语或政治思想方面与人工智能接触，即使政治对民主、公平和伦理的影响得到承认（例如参见 Shaw，2017 年;Lee，2018 年;Nour，2019 年）。
Ideas spur or uphold institutional and political changes (Blyth, 2002; Fairclough, 2010; Parsons, 2003; Pierson, 1992), reflect power and ideological struggles (Fairclough, 2010; Laclau and Mouffe, 2001), and affect both social and material reality (Schmidt, 2017). Following a long tradition in political science, ideas are recognised as the bases of ideologies and policy outcomes (regulations, policies, implementation, etc.). Ideas have real impacts on the conduct of social relations, political arrangements, and economic structures (Blyth, 2002; Gofas and Hay, 2010; Lindberg, 2018; Brink and Metze, 2006). Ideas can lock in specific policy pathways (Lee, 2018) and create the backbone of institutions that will provide stability (Blyth, 2002). Ideas can also, conversely, spur policy and institutional change. Blyth explains that ideas help interpret goals, causes, courses of action and actors involved in times of uncertainty, and enable collective action and coalition building. Finally, ideas can be ‘weaponised’ to change policy outcomes and institutions (Blyth, 2002). Schmidt (Schmidt, 2008) argues that institutions can be changed or maintained through idea mediation (Schmidt, 2017, 2008).
思想刺激或支持制度和政治变革（Blyth，2002 年;费尔克拉夫，2010 年;帕森斯，2003 年;Pierson，1992 年），反映了权力和意识形态斗争（Fairclough，2010 年;Laclau 和 Mouffe，2001 年），并影响社会和物质现实（Schmidt，2017 年）。遵循政治学的悠久传统，思想被认为是意识形态和政策成果（法规、政策、实施等）的基础。思想对社会关系、政治安排和经济结构的行为有真正的影响（Blyth，2002 年;Gofas 和 Hay，2010 年;林德伯格，2018 年;Brink 和 Metze，2006 年）。想法可以锁定特定的政策路径（Lee，2018），并创建提供稳定性的机构支柱（Blyth，2002）。相反，想法也可以刺激政策和制度变革。布莱斯解释说，思想有助于解释目标、原因、行动方案和不确定时期参与的行为者，并实现集体行动和联盟建设。最后，思想可以被“武器化”以改变政策结果和制度（Blyth，2002）。施密特（Schmidt，2008）认为，制度可以通过思想调解来改变或维持（Schmidt，2017,2008）。
Considering the many ideas on the risks and promises of AI and its impact on sustainability action and governance (Crawford, 2021; Dauvergne, 2021), scrutinising AI as a discursive referent in international sustainability policies is timely.
考虑到关于人工智能的风险和前景及其对可持续发展行动和治理的影响的许多想法（Crawford，2021 年;Dauvergne，2021 年），将人工智能作为国际可持续发展政策的话语参考对象进行审查是及时的。

2.2. Method  2.2. 方法

2.2.1. Case selection  2.2.1. 案例选择

International organisations hold critical potential as knowledge brokers, coordinators, agenda setters and programme implementers (Karns et al., 2015; Barnett and Duvall, 2004). They can redefine interests and worldviews for other international actors such as states (Barnett and Finnemore, 2004) and are “international social institutions characterized by behavioural patterns based on international norms and rules, which prescribe behavioural roles in recurring situations that lead to a convergence of reciprocal expectations” (Rittberger et al., 2019). They often serve as platforms or advocates in spreading and shaping new discourses and norms (Karns et al., 2015; Fukuda-Parr and Hulme, 2011), and as such are an important link in policymaking, ideational and institutional change.
国际组织作为知识经纪人、协调者、议程制定者和计划实施者具有重要潜力（Karns 等人，2015 年;Barnett 和 Duvall，2004 年）。它们可以重新定义国家等其他国际行为体的利益和世界观（Barnett 和 Finnemore，2004 年），并且是“以基于国际规范和规则的行为模式为特征的国际社会机构，这些行为模式规定了在反复出现的情况下的行为角色，导致相互期望的趋同”（Rittberger et al.，2019）。他们经常充当传播和塑造新话语和规范的平台或倡导者（Karns 等人，2015 年;Fukuda-Parr 和 Hulme，2011 年），因此是政策制定、意识形态和制度变革的重要环节。
This paper focuses on is the European Union (EU), the United Nations (UN) and the World Economic Forums in their role as ‘idea entrepreneurs’, i.e. as organisations or sets of organisations defending and promoting certain ideas in order to translate them into policies and norms (Levinson, 2021). Other international organisations, most notably the
本文重点关注欧盟 （EU）、联合国 （UN） 和世界经济论坛作为“思想企业家”的角色，即作为捍卫和推广某些想法的组织或组织集，以便将其转化为政策和规范（Levinson，2021）。其他国际组织，尤其是
Organisation for Economic Co-operation and Development (OECD), place a considerable focus on how AI principles are being formed (Schmitt, 2022). To delimit the study, we selected the three organisations because they represent distinct mandates and diverse types of international organisations. The UN is global intergovernmental with global reach mandated to spur sustainable development which gathers diverse types of economies and geopolitical interests. The EU is supranational, regional, and political with clear stakes in the development of AI. The World Economic Forum is non-governmental with significant convening power of business and government representatives and has a substantial analysis function. We chose organisations with distinct mandates to potentially capture different governance aspects and interests (Ulnicane et al., 2022; Schmitt, 2022).
经济合作与发展组织 （OECD） 非常关注人工智能原则的形成方式（Schmitt，2022）。为了界定研究范围，我们选择了这三个组织，因为它们代表了不同的任务和不同类型的国际组织。联合国是全球政府间组织，其全球影响力旨在促进可持续发展，从而汇集了各种类型的经济体和地缘政治利益。欧盟是超国家的、区域性的和政治性的，在人工智能的发展中具有明显的利害关系。世界经济论坛是非政府组织，具有重要的企业和政府代表召集权，并具有重要的分析职能。我们选择了具有不同任务的组织，以潜在地捕捉不同的治理方面和利益（Ulnicane 等人，2022 年;Schmitt，2022 年）。

2.2.2. Document selection and analysis
2.2.2. 文档选择和分析

Our corpus of documents is composed of policy documents, white papers and reports linking AI applications to sustainability. The corpus was found through searches on the official websites of the EU, the UN and the World Economic Forum up to February 2022. We used different combinations of the terms “AI”, "sustainab* ", "environment* ", “green” and “SDG”. We also included documents from Smart Africa, Microsoft, the International Research Center on Artificial Intelligence, and Future Earth for triangulation. We retained 28 documents altogether, published from 2018 to 2021.
我们的文件语料库由将人工智能应用与可持续发展联系起来的政策文件、白皮书和报告组成。该语料库是通过在欧盟、联合国和世界经济论坛的官方网站上搜索找到的，截至 2022 年 2 月。我们使用了“AI”、“sustainab*”、“environment*”、“green”和“SDG”等术语的不同组合。我们还包括来自 Smart Africa、Microsoft、国际人工智能研究中心和 Future Earth 的三角测量文件。我们总共保留了 28 份文件，从 2018 年到 2021 年发布。
After data selection, the analysis proceeded iteratively. Preliminary categories included metaphors, categorisations, problem descriptions, areas of implementation, and storylines. Additional subcategories were added through a continuous process of adjustment and reorganisation.
数据选择后，分析迭代进行。初步类别包括隐喻、分类、问题描述、实施领域和故事情节。通过不断的调整和重组过程增加了其他子类别。

2.3. Analytical framework
2.3. 分析框架

2.3.1. VDP triad  2.3.1. VDP 三元组

This paper focuses on political ideas. They are claims and assumptions about the desirable, possible, effective and legitimate goals and courses of action in society (Blyth, 2002; Lindberg, 2018; Schmidt, 2008; Vernon, 2010). Political ideas delimitate the ‘proper’ course of action, as well as its scope, place and time. Additionally, ideas may identify the agents and objects to be included, and present underlying assumptions about the world within which agents and institutions operate (Blyth, 2002; Schmidt, 2008; Lynch, 2007).
本文重点关注政治思想。它们是关于社会中理想、可能、有效和合法的目标和行动方案的主张和假设（Blyth，2002 年;林德伯格，2018 年;施密特，2008 年;Vernon，2010 年）。政治思想界定了“正确”的行动方针，以及它的范围、地点和时间。此外，想法可以识别要包含的代理和对象，并提出有关代理和机构运作的世界的基本假设（Blyth，2002 年;施密特，2008 年;林奇，2007 年）。
Political ideas are the substantive content of discourses (Schmidt, 2017, 2008). In this paper, we refer to discourse as “a specific ensemble of ideas, concepts and categorisations that are produced, reproduced, and transformed in a particular set of practices and through which meaning is given to physical and social realities” (Hajer, 1997).
政治思想是话语的实质内容（Schmidt，2017,2008）。在本文中，我们将话语称为“在一组特定的实践中产生、复制和转化的思想、概念和分类的特定集合，并通过这些概念和分类赋予物理和社会现实意义”（Hajer，1997）。
We use idea analysis and the VDP-triad to operationalise our analysis (Lindberg, 2018). The triad corresponds to (V) values or value-judgements; (D) descriptions or judgements of reality, i.e. problem formulation; and ( P ) ( P ) (P)(\mathrm{P}) prescriptions or practical proposals for actions.
我们使用想法分析和 VDP 三元组来实施我们的分析（Lindberg，2018）。三元组对应于 （V） 值或价值判断;（D） 对现实的描述或判断，即问题表述;以及 ( P ) ( P ) (P)(\mathrm{P}) 处方或实际行动建议。
We use the VDP-triad to identify how the governance aspects of AI are framed. Framing an issue involves defining problems, diagnosing causes, making moral judgments, and suggesting remedies in an internally logical manner (Entman, 1993; Wibeck and Linnér, 2021). We see frames as assemblages of ideas underpinning discourses. Framing analysis goes beyond the VDP-triad to also call attention to what is excluded.
我们使用 VDP 三元组来确定 AI 的治理方面是如何构建的。构建问题涉及定义问题、诊断原因、做出道德判断以及以内部逻辑方式提出补救措施（Entman，1993 年;Wibeck 和 Linnér，2021 年）。我们将框架视为支撑话语的思想组合。框架分析超越了 VDP 三元组，还引起了人们对被排除的内容的关注。

2.3.2. Sustainability discourses
2.3.2. 可持续发展论述

Our study presupposes that AI as a discursive referent will either instigate discursive change or reinforce already existing institutional and discursive designs. To explore these options, we provide a summary of three dominating sustainability discourses. The discourses used as reference are ecological modernisation (Hajer, 1997; Bäckstrand and Lövbrand, 2006), green governmentality (Bäckstrand and Lövbrand, 2006, 2019; Rutherford, 2017; Luke, 1995), and civic environmentalism (Bäckstrand and Lövbrand, 2006, 2019). These three discourses stem from the Enlightenment perspective. Other discourses such as
我们的研究假设人工智能作为话语所指对象将引发话语变革或强化已经存在的制度和话语设计。为了探索这些选择，我们总结了三种主要的可持续发展话语。用作参考的话语是生态现代化（Hajer，1997 年;Bäckstrand 和 Lövbrand，2006 年）、绿色政府（Bäckstrand 和 Lövbrand，2006 年、2019 年;卢瑟福，2017 年;Luke，1995）和公民环保主义（Bäckstrand 和 Lövbrand，2006 年，2019 年）。这三种话语源于启蒙运动的观点。其他话语，例如
environmental justice or survivalism also exist (Zannakis, 2009), but they did not appear in any significant manner in our corpus.
环境正义或生存主义也存在（Zannakis，2009），但它们并没有以任何重要的方式出现在我们的语料库中。
Ecological modernisation is understood as “the discourse that recognizes the structural character of the environmental problematique but nonetheless assumes that existing political, economic, and social institutions can internalize the care for the environment” (Hajer, 1997). Hajer explains that ecological modernisation (1) makes sense of environmental change in a way that is calculable, from a monetary perspective in particular; (2) frames environmental changes as a positive-sum game enabled by the cooperation between different actors; and (3) assumes that economic growth and environmental protection are coupled. This discourse favours small tweaks in politics and the economy in the form of green, decentralised, cost-effective and collaborative regulations, technological innovation, investment and trade (Bäckstrand and Lövbrand, 2006).
生态现代化被理解为“承认环境问题的结构性特征，但仍然假设现有的政治、经济和社会机构可以内化对环境的关怀的话语”（Hajer，1997）。Hajer 解释说，生态现代化 （1） 以一种可计算的方式理解环境变化，特别是从货币角度;（2）将环境变化视为由不同行为者之间的合作促成的正和博弈;（3）假设经济增长和环境保护是耦合的。这种论述倾向于以绿色、分散、具有成本效益和协作的法规、技术创新、投资和贸易的形式在政治和经济方面进行小调整（Bäckstrand 和 Lövbrand，2006 年）。
Green governmentality focuses on how governmental mechanisms and technologies are used to govern environmental problems (Rutherford, 2017; Fletcher, 2017). It also addresses how knowledge of environmental processes is construed, negotiated, and represents sites of power production (Rutherford, 2017). Production of knowledge about the environment steers the “management” of the natural world and individuals’ behaviour. Put simply, the environment is (re)presented by means of statistics, maps and models through which states, international and non-governmental organisations can devise solutions (Rutherford, 2017). This is conducive to techno-solutionism that construes the planet and its ecosystems as “infrastructure [ s s ss ] subject to state protection, management and domination” (Bäckstrand and Lövbrand, 2006, p. 55). The way the environment is made knowable and governable often leads to the creation of specific types of self-governing subjects, for example the green citizen (Rutherford, 2017).
绿色政府侧重于如何使用政府机制和技术来管理环境问题（Rutherford，2017 年;弗莱彻，2017 年）。它还讨论了环境过程知识如何被解释、协商和代表电力生产地点（Rutherford，2017）。关于环境的知识的产生引导着自然世界和个人行为的“管理”。简而言之，环境是通过统计数据、地图和模型（重新）呈现的，国家、国际和非政府组织可以通过这些模型制定解决方案（Rutherford，2017）。这有利于技术解决方案主义，将地球及其生态系统解释为“受国家保护、管理和统治的基础设施” s s ss （Bäckstrand 和 Lövbrand，2006 年，第 55 页）。环境变得可知和可治理的方式通常会导致特定类型的自治主体的产生，例如绿色公民（Rutherford，2017）。
Civic environmentalism stipulates that an effective way to deal with environmental problems is through increased participation in the decision-making process (Bäckstrand and Lövbrand, 2006). All those
公民环保主义规定，处理环境问题的有效方法是增加对决策过程的参与（Bäckstrand 和 Lövbrand，2006 年）。所有这些
affected (minorities and women in particular) by environmental change should have the opportunity to have their voices heard. Bottom-up approaches are to be favoured. However, the extent to which political structures should be reformed and civic engagement secured varies between more reform-oriented and radical approaches to civic environmentalism (Bäckstrand and Lövbrand, 2006).
受环境变化影响的人（特别是少数群体和妇女）应该有机会听到他们的声音。自下而上的方法应受到青睐。然而，政治结构的改革和公民参与的保障在多大程度上有所不同，在更注重改革和激进的公民环保主义方法之间（Bäckstrand 和 Lövbrand，2006）。

3. Key idea frames
3. 关键思想框架

We identified two frames. They mediate the understanding of AI as a mode of sustainability governance -i.e. as ways to develop and implement regulations and incentives in order to change behaviours (O’Neill, 2017)- and as an object of governance in its own right. One frame represents AI as a tool for sustainability governance and the other advocates for a governance framework to promote sustainable AI. We abridged the first frame as AI for sustainability and the second as Governance of sustainable AI. Fig. 1 provides a summary of the results. In the case of AI policies, the term governance is rarely defined (Ulnicane et al., 2022). We define governance as an ensemble of rules and networks of actors and practices aiming to manage political problems.
我们确定了两个框架。它们调解了对人工智能作为一种可持续性治理模式的理解——即作为制定和实施法规和激励措施以改变行为的方式（O'Neill，2017）——以及作为其本身治理对象的理解。一帧表示人工智能是可持续发展治理的工具，另一帧则倡导建立治理框架以促进可持续人工智能。我们删节了第一个框架作为可持续发展的人工智能，第二个框架作为可持续人工智能的治理。图 1 总结了结果。就人工智能政策而言，治理一词很少被定义（Ulnicane 等人，2022 年）。我们将治理定义为旨在管理政治问题的一系列规则和行为者和实践网络。
Although “AI” is an umbrella term that encompasses a variety of technologies with different applications, opportunities and challenges, the framing of AI is not always specific in the documents selected. We therefore kept AI as a general referent for the analysis.
尽管“人工智能”是一个总称，涵盖了具有不同应用、机遇和挑战的各种技术，但在所选文件中，人工智能的框架并不总是具体的。因此，我们保留人工智能作为分析的一般参考对象。

3.1. AI for sustainability
3.1. 人工智能促进可持续发展

3.1.1. Values  3.1.1. 价值观

The UN, the EU and the World Economic Forum explicitly state that the use of AI should be steered in a way that contributes to the Sustainable Development Goals (SDGs) (World Economic Forum, 2018; European Commission, 2021a; UN Global Pulse, 2019, 2018, 2020; International Telecommunication Union, 2020; Chief Executive Board for
联合国、欧盟和世界经济论坛明确指出，人工智能的使用应以有助于实现可持续发展目标 （SDG） 的方式引导（世界经济论坛，2018 年;欧盟委员会，2021a;联合国全球脉搏，2019 年、2018 年、2020 年;国际电信联盟，2020 年;行政总裁
Fig. 1. Summary of the frames found.
图1.找到的帧的摘要。
    • Corresponding author at: Department of Thematic Studies, Environmental Change, Linköping University, Sweden.
      通讯作者：瑞典林雪平大学环境变化专题研究系。
    E-mail address: marie.francisco@liu.se (M. Francisco).
    电子邮件地址：marie.francisco@liu.se （M. Francisco）。
    1 1 ^(1){ }^{1} Address: Campus Valla, TEMA-huset, Ingång 37, Rum E:243, Linköping Universitet, 58183 Linköping
    1 1 ^(1){ }^{1} 地址：Campus Valla， TEMA building， Entrance 37， Room E：243， Link 雪 ping University， 58183 Link 雪平