Konferenz-Abstract
Assessing Sustainable Artificial Intelligence via Societal Impact Analysis: The Case of Earth Observation
Vorschaubild nicht verfügbar
Volltext URI
Dokumententyp
Text/Conference Abstract
Zusatzinformation
Datum
2023
Zeitschriftentitel
ISSN der Zeitschrift
Bandtitel
Quelle
Verlag
Gesellschaft für Informatik e.V.
Zusammenfassung
In the pursuit of fighting the climate catastrophe several technologies promise to help creating sustainable societies. In the digital realm a supposedly revolutionary set of technologies called artificial intelligence (AI) is currently discussed as tool for reaching ecological sustainability. Ecologically oriented AI applications like data analysis and monitoring as well as decision automation have been the subject of research for some time, and recently, the ecological footprint of AI systems themselves started to come into focus too.
On a different note and shifting the focus to social sustainability, the societal implications of AI applications directly affecting individuals and groups are widely discussed in terms of fairness, accountability and transparency as well as regarding ethics and regulation. However, ecologically oriented AI applications also do have societal implications, although they are often less obvious, partly because of the ecological focus and purpose, and partly because of the usually more indirect relation. This paper intents to close this gap by suggesting an appropriate general societal impact analysis grid and by applying this grid on the concrete use case of satellite-data driven Earth Observation (EO).
This approach is fruitful in two ways: First, our suggested societal impact analysis grid extends existing technology impact assessment frameworks (like the rather general Matrix of Convivial Technology) with critical social theory aspects (e.g. fairness, inequality), and data protection theory aspects (e.g. information power asymmetries). With this extension the grid is especially applicable to societal implications of computational and data-related technologies like AI. Second, we exemplify the use of our analysis grid on the case of EO and therefore uncover societal implications not evident upon first sight.
Within its purpose EO provides practical benefits illustrating the potential of AI in contributing valuable sustainability-related insights, e.g. environmental monitoring. However, applying our grid, societal issues do emerge, e.g. a critical observer-observed relationship or global structural unfairness concerning available data. In creating the impact grid and discussing the concrete case of EO, this paper contributes to a needed and growing scholarship of sustainable AI.