Sustainability in a digital world

Late last year, the World Wide Web Consortium (W3C) – the same organisation responsible for the Web Content Accessibility Guidelines with which all websites in the UK are legally required to comply – published a draft note detailing multiple layers of guidance on what they believed could help achieve a more sustainable web. This draft note, the Web Sustainability Guidelines (WSG), was meant for a diverse audience involved in the lifecycle of creating digital products, from digital teams to policymakers and those involved in running the business. This was innovative in that, for the first time, actionable guidance to help digital teams understand sustainability in technology was being articulated. It went beyond just the technology hardware and software themselves, but also included the entire process required to run it – from the business aspect to the social and finance considerations and more. 

There is increasing awareness that digital technology and AI have negative impacts on the environment. In the initial phase when ChatGPT was first launched, well-intentioned people started estimating the amount of water consumed by each AI query or image generated. However, because the Internet is distributed in different geographical locations all over the world, with endless possibilities of combinations of hardware and software, and high variability in consumer usage patterns, it is nearly impossible to estimate even ballpark figures that may be representative of the true environmental costs. In other words, environmental costs of technology use is highly context-dependent, hence it is difficult to prescribe a blanket solution beyond high-level principles for the sustainable use of technology.  

People use technology in legitimately beneficial ways that improve their lives. For example, we rely on quick Google searches to find information we need, and students whose first language is not English rely on AI translation to understand difficult concepts that they are learning. The resulting cognitive dissonance between the benefits and environmental impacts of technology and AI is real, and can engender a sense of helplessness, anxiety, and guilt in people who genuinely want to try to reduce their carbon footprint. 

For some time now, in a corner of the web, there has been a small community of people passionate about making the web more sustainable. Amongst some of their experiments was a grid-aware website that could adapt your website to be less carbon intensive if the grid your device was running on was not as clean; another was a calculator that estimated how much emissions your website was producing. Through these experiments and more, they have grappled with the ambiguity and uncertainties of measuring the environmental costs of digital technology mentioned earlier. The result is WSG as an articulation of what a sustainable web could look like, which has been useful for Imperial to adapt and adopt for its digital estate and the page views it serves. 

Imperial’s first WSG, published on 13^th March this year, is a first step to try to reduce the environmental impact it takes to deliver Imperial’s digital content. It is meant to guide content editors, regardless of their technical background, on decisions they can control, such as what content is published, how pages are structured and formatted, and how media assets are used. Although each web page’s emissions seem minute at 3 to 5mg of CO₂, the University of Edinburgh’s implementation of the sustainable web design principles across more than 65,000 webpages has shown to achieve potential emissions reduction of 25 tonnes annually, so there could be non-trivial savings with Imperial’s similarly-sized digital estate. The principle of compounding, well-known in finance to produce exponential growth in savings, likewise applies when it comes to emissions savings. 

Some of the guidance poses co-benefits, such as improving user experience, accessibility and performance. However, there are also real trade-offs involved. For example, reducing the use of images, videos, and animations reduces energy consumption, but may also reduce the site’s visual appeal. More importantly, what is valuable is the process of different parties coming together to assess and consider the environmental trade-offs, and building a culture of giving the planet a seat at the table in decision-making. 

Ultimately, what is digital requires physical resources to power, but these environmental costs tend to be out of sight and out of mind. Since web sustainability principles can enable emissions reductions at source, they will be increasingly important to consider as one of many tools to reduce technology’s energy footprint amidst growing demand for digital services, including AI.

Thanks goes to Susie Yang who worked on the drafts of Imperial’s web sustainability guidelines, and Yilong (Calvin), who worked on the tools, calculations, and analyses for web sustainability.