Convivial AI? Developing a societal impact analysis grid for assessing artificial intelligence in Earth observation

The deployment of artificial intelligence (AI) applications with an ecological focus for data collection, analysis, monitoring, and decision automation has been widely described as a promising way of achieving sustainability. Such use has been the subject of research for some time, and recently, the ecological footprint of AI systems themselves has also been considered. While the societal implications of common AI applications are widely researched in terms of fairness, accountability, and transparency, the societal impact of specialized, ecologically-oriented AI applications remains understudied. To address this gap, we designed an analysis framework – the societal technology impact assessment grid (STIAG) – by extending the matrix of convivial technology (MCT), thus making it suitable for analyzing AI applications. We then apply this framework to satellite data-driven Earth observation (EO). This article seeks to communicate insights on two fronts: (1) the STIAG extends existing sustainability-oriented technology impact assessment frameworks, such as the MCT, by incorporating critical social and data-protection theory and (2) we contribute to the EO research domain by applying the analysis grid to EO’s AI-based methods, thereby uncovering their societal – and indeed neo-colonial – implications. We furthermore aim to advance the growing body of critical scholarship on sustainable AI and the field of EO itself by providing a constructive foundation for sustainable EO practices that can also inform research design and policymaking.

以生态为导向、用于数据采集、分析、监测与决策自动化的人工智能（artificial intelligence, AI）应用部署，已被学界广泛认为是实现可持续发展的极具前景的路径。此类应用的研究已有一段时日，近来AI系统自身的生态足迹也逐渐受到关注。尽管学界已围绕通用AI应用在公平性、问责制与透明度层面的社会影响展开了大量研究，但针对专业化生态导向AI应用的社会影响仍有待深入探索。为填补这一研究空白，我们通过拓展共生技术矩阵（matrix of convivial technology, MCT），设计了一套适用于AI应用分析的分析框架——社会技术影响评估网格（societal technology impact assessment grid, STIAG）。随后我们将该框架应用于卫星数据驱动的地球观测（Earth observation, EO）场景中。本文旨在从两大维度阐明研究发现：(1) 本框架通过融入批判社会与数据保护理论，拓展了以可持续发展为导向的现有技术影响评估框架（如共生技术矩阵MCT）；(2) 我们将该评估网格应用于地球观测领域的AI驱动方法中，揭示了其社会层面乃至新殖民主义层面的潜在影响。此外，本文旨在为可持续地球观测实践搭建建设性基础，以此推动可持续人工智能领域及地球观测领域日益壮大的批判学术研究发展，同时为相关研究设计与政策制定提供参考。