AI with human touch needed to reduce bias in climate planning

Artificial intelligence

Artificial intelligence used to plan and mitigate climate change can be skewed if incomplete data sets are used – and a human touch is needed to redress the balance, say researchers.

AI climate science computer programmes pick through complex datasets looking for patterns and insight.

But information can be missing from places in the world or time periods. There can also be other critical data missing such as social information, markets, and infrastructure.

Patchy information can create bias and result in poor weather predictions, or an underestimate of carbon emissions from industry.

This could mislead climate change action and lead to further harm to those, particularly in the global south, more vulnerable to the extreme weather events.

Primary author and fellow at Cambridge Zero (a University of Cambridge initiative to research ways to mitigate climate change) Dr Ramit Debnath said: “When the information on climate change is overrepresented by the work of well-educated individuals at high-ranking institutions within the global north, AI will only see climate change and climate solutions through their eyes.”

Human-in-the-loop design

Researchers argue that using a ‘human-in-the-loop’ design would ensure biases were noticed and addressed. The design allows for a sense check of the data and context. They argue this can help fill in the gaps to improve the accuracy of predictions.

The paper, published in Nature’s npj Climate Action series, highlighted ChatGBT as it can ask its human users follow up questions, and challenge and admit mistakes.

Researchers also argued that broadband should be seen as a public necessity rather than a private commodity. Broadband is key to improving the data we have and address data gaps, as well as encourage more people to take part in conversations around AI and climate change.

Co-author, Cambridge Zero director and climate scientist Professor Emily Shuckburgh, said: “No data is clean or without prejudice, and this is particularly problematic for AI which relies entirely on digital information.”

Shuckburgh leads the UK national research funding body’s (UKRI) Centre for Doctoral Training on the Application of AI to the study of Environmental Risks (AI4ER). She added: “Only with an active awareness of this data injustice can we begin to tackle it, and consequently, to build better and more trustworthy AI-led climate solutions.”

Don’t miss out on BIM and digital construction news: sign up to receive the BIMplus newsletter.

Story for BIM+? Get in touch via email: [email protected]

Latest articles in News