Algorithms in the room: AI, representation, and decisions about sustainable futures

This article considers the role of generative AI technologies, such as large language models (LLMs), in promoting the views of underrepresented groups. We are specifically concerned with the role AI could play in encouraging powerful decision-makers—often leading politicians and businesspeople in Western nations—to consider the perspectives of underrepresented groups when making decisions about sustainable development. Some suggest generative AI could offer decision-makers perspectives they had previously not considered, leading to more equitable and innovative policy approaches, and supporting several of the United Nations' Sustainable Development Goals (SDGs). We critique this perspective. Groups may be underrepresented in sustainable development decision-making because of individual cognitive and organisational information-processing limitations (‘ omitted, but not opposed ’), and because of opposition which remains even if these limitations are overcome (‘ opposed, whether omitted or not ’). We outline how these ‘categories of omission’ shape the opportunities and risks created by generative AI in representative sustainability. • Sustainable development requires involvement from many different stakeholders. • Some stakeholders are underrepresented due to neglect or political opposition. • Generative AI may support underrepresented groups in several ways. • Generative AI may exacerbate underrepresentation given existing opposition.