When it comes to AI, much of the discussion within universities focuses on student use of generative AI tools and the energy demands of large-scale language models. However, this narrow focus risks overlooking important parts of the bigger picture. Academics themselves are increasingly using AI in their research.
From collecting and managing large datasets, to generating synthetic data, analyzing data with machine learning techniques, and using LLM to support literature review, coding, and writing, AI is becoming embedded in everyday academic research.
Can artificial intelligence be as environmentally friendly as it is essential to solving complex problems? As universities integrate courses on AI into different disciplines, they also need to show students the risks and costs, especially to the environment.
At our university, environmental researchers are conducting research on how to reconcile the use of AI with values. We found that while researchers generally recognize the impact that AI will have on the environment, it rarely leads to real, lasting change. This was surprising given researchers’ commitment to environmental values and efforts to address environmental issues, so we investigated what was going on.
While a few researchers sought to reduce their environmental footprint by using small, locally hosted models, most researchers managed the tension between AI methods and environmental research through various strategies that shifted responsibility for the impacts of their own AI use elsewhere.
Most commonly, responsibility was shifted upwards to universities, funders, and institutional systems. Decisions about cloud storage, computing infrastructure, and access to servers were considered to be beyond the control of individual researchers, but increasing pressure to publicly demonstrate productivity has made it difficult to prioritize slower, less computationally intensive methods.
Responsibility was also transferred Landscape To other researchers who may be working with large-scale, resource-intensive AI systems. However, what counts as a larger model remains largely undefined, blurring the line between acceptable and excessive AI use.
Someone else passed the blame on. further awaypoints to technology companies, data centers, and global supply chains as the real culprits of environmental damage. Some researchers looked to policymakers to address the issue, while others framed AI’s environmental impact as a future problem to be addressed when better technology, more efficient systems, or stronger governance emerge.
A combination of these blame-shifting methods has allowed researchers to continue using AI while avoiding the uncomfortable knowledge that its use is having negative environmental effects.
However, our findings suggest that it is neither realistic nor desirable to place responsibility for these effects solely on individual researchers. Researchers occupy the tail end of the computing supply chain and have little influence over the design of AI systems, the procurement of energy, the procurement of computing infrastructure, and the institutional incentives that shape research practices.
In contrast, university leaders teeth We are in a position to take responsibility for the environmental impact of research using AI. If we are serious about reducing the environmental impact of research, responsibility needs to go beyond individual choices and be addressed at an organizational level.
We suggest four actions universities can take to reduce the environmental impact of research using AI.
1. Measure and visualize impact
University leadership teams should integrate carbon and computing tracking tools into research computing systems to provide dashboards that help researchers, departments, and institutional leaders understand the environmental impact of their work.
Environmental costs are often invisible and therefore often overlooked. Measuring and reporting these impacts is a critical first step toward informed decision-making and organizational accountability.
2. Default to low-impact options
Researchers who actively sought to reduce the environmental impact of our research often had to do so on their own initiative. Universities can facilitate sustainable choices by providing access to small-scale open source models, locally hosted AI systems, and energy-efficient shared computing infrastructure. Adding AI should not be a default option in the provided software. Procurement policies should also prioritize providers who are transparent about their energy use and emissions. The goal is to make low-impact approaches the easiest and most accessible options, rather than allowing researchers to make these choices alone.
3. Incorporating AI sustainability into research governance and incentives
The environmental impact of AI should not be placed in a separate AI policy silo. Instead, universities should embed consideration of the environmental impacts of AI into existing policies and processes, such as net zero, sustainability and climate change strategies, procurement decisions, research integrity frameworks, research ethics reviews, and data management plans.
Consider how funding, performance, and promotion criteria may unintentionally reward increasingly compute-intensive research. Building sustainability into governance and incentive structures ensures that it becomes a routine part of decision-making, rather than an optional add-on.
4. Establish clear organizational ownership
Responsibility for the environmental impact of AI is often distributed across research support teams, IT services, procurement offices, sustainability teams, and academic departments. Establish a cross-cutting group with the power to coordinate action and set policy across these areas. Without clear ownership, liability can easily slip through the cracks. This will enable universities to develop a consistent and responsible approach to sustainable AI.
The challenge is not simply to help researchers make better choices. Researchers are at the end of a complex computational supply chain and have limited influence over many of the factors that shape AI’s environmental footprint. If universities want researchers to use AI responsibly, they need to create institutional conditions that enable, make visible, and reward responsible use.
Sarah Hartley is Professor of Technology Governance and Director of the Center for Responsible Innovation at Exeter University Business School. Emily Robinson is an emerging technology, AI, society and sustainability researcher. Mayra Rodriguez is a Research Fellow specializing in AI and Data Science for Environmental and Health Applications at the University of Exeter.
Want advice and insights from academics and university staff delivered straight to your inbox every week? Sign up for campus newsletter.
