Knowledge is power. So is data

AI could help alleviate climate change. It could also make things worse

Wildfires in Maui. Floods in Italy. A cyclone in South East Africa. Different events in different locations all traceable to the same core problem: climate change. There’s a good chance you’ve heard the term recently. But what does it mean? The answer varies from region to region, country to country, and continent to continent. One of the most important differences in perspective is between the Global North (the world’s wealthiest countries) and the Global South (the world’s poorest countries).?

A key aspect of this difference is the level of contribution to climate change and the experience of its effects. An article on the website for Generation Climate Europe states that although the Global North contributes substantially more to climate change the Global South suffers more from its effects. And the North’s dominance of the global conversation impedes attempts to redress this balance. The BBC reports on a study examining the 100 most cited climate change research papers in a 5 year period. Fewer than 1% of the authors were from Africa and only 12 were female.????

We’re at risk of reinforcing this issue with AI. In an article on TheNextWeb Linnea Ahlgren describes a recent study which argues that although AI can be beneficial in the fight against climate change it risks reinforcing a view of the world that privileges the Global North over the Global South. The North has a far greater ability than the latter to record data and influence the development of technology. This means that AI solutions will disproportionately represent the North’s concerns and values.

This debate can be traced down to a regional level. The Generation Climate Europe article highlights that within countries, the worst effects of climate change are borne by ‘the poorest communities, including Black, Indigenous and People of Colour communities (BIPOC), as well as women and children’. It’s known that these communities are at risk from AI in multiple domains including but not limited to:?

• Policing - A tool for predicting likelihood to reoffend showed systematic bias against Black offenders
• Recruitment - A tool for selecting the best applicants for roles systematically discriminated against women
• Credit Scoring - A tool for assigning lines of credit gave lower limits to women than to men

There’s no reason to believe that AI applications for addressing climate change would be immune to the problem.

A few weeks ago I wrote that the decision about applying technology should depend on who will gain power. The climate change debate is the latest example of this question. Without a conscious effort to collect data in a way that gives proper representation to less wealthy communities we risk reinforcing power imbalances and making the problem worse.

AI could be used to somewhat even the playing field between Global North and South. But without careful planning it will do the exact opposite.