Universities ‘should demand emissions data’ from AI firms

University leaders have been urged to lobby firms to release environmental data connected to the use of artificial intelligence to help institutions get a handle on how increased reliance on the technology is impacting on climate goals.

As concerns grow over the mostly hidden emissions generated by AI, UK IT body Jisc and the Environmental Association for Universities and Colleges (EAUC) have published a sustainability guide for the education sector. 

Unveiled as part of the annual EUAC conference, the report recommends university, college and sector leaders rally together to demand greater transparency from AI vendors on the technology’s water consumption and energy use.

“For institutions, the challenge is that energy consumption data for specific AI tools is rarely disclosed,” the guide states.

“Most vendors do not provide per-query or per-user energy figures, making it difficult to incorporate AI use into institutional carbon accounting.”

But by asking more of suppliers when it comes to transparency, leaders will “signal institutional priorities and contribute to growing demand for better disclosure”, authors suggest.

Intended to “highlight both the risks and opportunities of AI” for environmental good practice, the document calls on universities and other tertiary educational settings to conduct an audit of their AI use, and educate staff and students on the technology’s impact.

“None of these actions will transform an institution’s environmental footprint overnight,” the guide notes.

“But they will help to build the foundations for more substantive action and engagement as better data, standards, and regulatory frameworks emerge.”

Billed as a “practical guide”, AI and environmental sustainability in post-16 education seeks to illustrate how organisations can both take stock of their AI use and its environmental impact, as well as how AI can support sustainability ambitions, including by “accelerating” climate research and automating emissions reporting.

Another recommendation is that institutions educate staff and students on how to submit genuinely instructive directions to AI, to make it less likely machines will fail to produce useful answers. Prompting poorly, the guide suggests, contributes to wasteful emissions, and so “learning to prompt effectively is both a practical skill and an environmental one”.

Researchers point to several universities as examples of good work on the intersection between AI and the environment, with the University of Cambridge and University of Northampton among those spotlighted.

In one case study, the University of Wales Trinity Saint David is singled out for its creation of an AI ethics group, established “in direct response” to students’ environmental and ideological concerns about the technology.

“The group has been working through questions that many institutions are only beginning to grapple with, such as what students’ concerns are and how these differ between disciplines and demographics, how staff should respond when students refuse AI on ethical grounds, whether assessments should ever mandate AI use, and how to support informed decision-making rather than simply prescribing tools,” the report notes.

In another example, the University of Oxford is credited with pioneering work measuring AI use across its estate. The institution’s gardens, libraries and museums division is currently piloting a “digital sustainability measurement tool” to quantify staff’s reliance on generative AI specifically, in an attempt to build a “clearer picture of demand and associated environmental impacts”.

Co-author Charlotte Bonner, the CEO of EAUC, told Times Higher Education (THE) that the partnership between EAUC and Jisc was about “wanting to raise people’s confidence” when it comes to interacting with AI.

“What we were hearing from our members is ‘we know this is a thing, we know that they’re connected, but we don’t have a trusted source of information that we can easily navigate’,” said Bonner, of the technology and its relationship to the environment.

Her co-author, subject specialist in digital sustainability at Jisc, Cal Innes, added: “It’s very much just the rate of change...We’re talking about such a recent phenomenon. It’s only been three years since ChatGPT and the AI boom…I think everybody really is just grappling to try and make sense of all the noise.”

Asked whether technology suppliers might simply ignore demands for transparency on emissions, Bonner suggested the education sector had more playing power than one might imagine.

“The collective purchasing power of the universities and post-16 sector is not insubstantial...There is a big marketing interest in the student base,” Bonner said.

georgia.luckhurst@timeshighereducation.com