The environmental harms of AI and the digital: Implications for education policy and research

• In Event: Education policy: actors, ideas and implications

Proposal

In recent years, there has been massive growth in the use of artificial intelligence (AI) and other digital technology across sectors, including in education and education policy making. Tailing behind this, but also now picking up speed are questions about the social and environmental harms of this technology use. Discussions of equity and other social repercussions have been more extensive, such as racial and cultural bias built into algorithms used in governance, or tech industry influence on public sectors through a growing reliance on corporate digital platforms (Gulson, Sellar, & Webb, 2022). This review of the literature takes focus on the yet lesser developed area of research on the environmental harms of AI and other digital technology. Through a systematic approach to reviewing literature across disciplines, it aims to provide a state of research in this nascent area, with a particular focus on implications for research and policy in the education sector.

To date, when the environment is linked to digital technology in research or policy spheres, it is typically in extolling the possibilities of technology for meeting climate and other environmental challenges, including in education. For example, a recent UNESCO Digital Learning Week event centred on the ‘dual challenges of the digital and green transitions’ (UNESCO, 2024). However, papers focused largely on ‘the potential’ of the digital for greening education, such as in ‘combating climate change through digital tools,’ or AI as a ‘pathway to a sustainable future.’ Not in view were the extensive environmental and environmental justice costs of the digital, which are just beginning to be tallied. For example, running a single natural language processing model has been estimated to have the same lifetime emissions as five cars (Crawford, 2021). However, challenges to understanding this and other environmental harms include the lack of access to information held by private tech corporations, such as figures on AI water or energy consumption. However, several recent book length studies have begun to explore the associated costs and considerations, such as ‘The Atlas of AI’ (Crawford, 2021) and ‘Is AI Good for the Planet’ (Brevini, 2022). The broader state of empirical research and academic study of this intersection is as of yet minimal and distributed across a range of academic disciplines.

In education, there has been very little work to date on the intersection of environment and the digital, consisting mainly of conceptual papers and blog posts (e.g., McKenzie & Gulson, 2023; Selwyn, 2022). However, AI and digital engagement in education provision and governance is also significantly on the rise, from the use of virtual tutors (Afzal et al., 2019) and AI teaching assistants (Baker & Smith, 2019), to the use of natural language processors like ChatGPT in the classroom (Galindo-Domínguez et al. 2024; Hays, Jurkowski, and Sims 2024), to the use of Google Classroom, Amazon Web Services (AWS), and other platforms (AWS 2024; Google for Classroom n.d.). Frameworks for assessing and decision-making on the extent and type of AI/digital use in schooling are as a result becoming more common (e.g., UNESCO, 2024). Participatory and collaborative forms of policy making are often being engaged in these processes, including to bring the technical expertise to bridge with education expertise in such sociotechnical decision-making (Callon, Lascoumes & Barthes, 2011; Thompson et al., 2022). However, while social and equity expertise and considerations of AI and the digital are increasingly raised in these forums, there remains a gap in bringing environmental and environmental justice considerations to the table, despite also increasing climate and sustainability priorities in education systems (McKenzie & Gulson, 2023).

Through this paper we aim to help address these gaps through providing an overview of the state of current research understandings on i) the environmental implications of AI/the digital, including in education; and ii) some implications of this for education policy decision-making. The paper will begin with an overview of the methods used in the review, including databases and search terms used. Study findings include breakdowns of the existing research on the intersection of environmental and digital priorities in schooling by environmental issues (e.g., mining, water use, energy consumption, greenhouse gas emissions, waste, environmental justice); by lifestage of technology development and use (e.g., extracting materials, building the digital infrastructures, or operating the AI or other forms of digital platforms and technologies used in schooling); and by prognoses or recommendations offered. Some recommendations in the literature include moving towards ‘‘green algorithm accountability’ (Brevini, 2022), or other shifts towards ‘green tech’ approaches. However, others are concluding that AI technologies may be ‘irredeemable’ (Selwyn, 2022) as a rapidly growing accelerator of an already quickly warming climate, and that more drastic approaches to limiting the digitalisation of education and society are needed. This paper will review this and other work on this issue that is of great relevance to the conference theme and education systems globally, with the session of interest to both researchers and policy makers.

Authors

• Marcia McKenzie, University of Melbourne

• Keara Quadros, Melbourne Graduate School of Education, University of Melbourne

• Kalervo Gulson, The University of Sydney