Approaches to environmental, social, and corporate governance — otherwise known as ESG — may be coming under increasing scrutiny, but it remains a boom industry. Ironically, the pursuit of socially conscious finance could lead to an alarming trend of so-called “data sweatshops” springing up across the globe to service the ESG industry itself. This should give pause for thought to those funnelling billions of pounds at the behest of so-called ethical investment.
The demand from companies to be scored for their ESG efforts has led to numerous data companies surfacing to provide ranking services, processing the vast amounts of data on supply chains and carbon footprints to produce companies’ ESG scores. The AI algorithms used by these services to calculate a sustainability rating rely on vast amounts of raw data, which in turn must be inputted manually through a labour-intensive process of “data labelling”. That is, labour which can mean that it ends up taking place in the same poor working conditions that ESG is supposedly addressing.
Data centres have appeared in less economically developed areas of the world, including parts of Africa, the Philippines and Indonesia, with data labellers working in insecure conditions on less than the minimum wage. Labellers add labels (i.e. amount of carbon emissions produced) in a code or language that the ESG score algorithm can recognise. This makes it possible for AI to identify a set of data, image or video. In the case of ESG, AI brings together large amounts of data with the intent on producing certain types of ESG scores which are displayed on funds being sold to investors, sometimes under the title of “Sustainability Characteristics” on the product information.
This information includes a cocktail of numbers, ranging from “carbon intensity” in tonnes to other metrics, such as one fund’s sustainability rating against others. These numbers indicate to the investor how ESG-friendly the company actually is, in the hope this will result in their receiving more funding, and deter support from less ESG-friendly organisations.
Other companies (mostly social media groups, who themselves are touted as “ESG” investments) might employ these sweatshops to do other labelling tasks as well, meaning data labellers could also be exposed to horrific content including bestiality and child sexual exploitation. They would be labelling images and videos for other clients, all the while being forced to work for as little as $1.32 per hour. Tanya Goodin, Tech Ethicist at the Royal Society of Arts, has characterised these places as not unlike “the disposable fashion industry and their sweatshops”.
The concept of ESG grew out of an increasingly morally conscious consumer movement, and arguably a political one, in the late 2000s to encourage investment into more socially conscientious companies. This push for a more “ethical capitalism” achieved early results, increasing the scrutiny on many companies who are using sweatshops and forced labour directly in the creation of their products or hidden deep in their supply chain. For example, Tony’s Chocoloney was forced to address the use of child labour in its supply chains, while tech firms have been less proactive in making changes in the face of pressure to do so.
But with demand only increasing for data and AI-driven ESG (as the World Economic Forum has called for), it will be increasingly hard for genuinely socially conscious investors to square the circle between the exploitation of those in the developing world and the lofty aims of ESG. If these efforts carry a charitable goal, then to do that we need to start reconsidering the practical real-world merits of ESG and whether it truly is the answer to more “ethical capitalism”.
When we’re swapping retail sweatshops for data sweatshops, it’s time to start questioning ESG.