Datafest, a recent event at The Shard in London, marked the culmination of a twelve week effort by students participating in the 2023 Data Science for Social Good (DSSG) Summer Fellowship programme at the University of Warwick, supported by the London Mathematical Laboratory.These activities are part of the larger Data Science for Social Good (DSSG) international programme, which originated at the University of Chicago in 2013. DSSG aims to equip data scientists with skills to tackle pressing societal issues. At Datafest, students reported on four specific projects which were undertaken this summer. In a previous blog post, we reported on one of these projects, focussing on improving predictions of deforestation in the Brazilian Amazon. Here we’ll offer a brief summary of progress made on another project aiming to help consumers to identify “greenwashing” in corporate advertising. Identifying green messaging and the potential for greenwashing In the context of climate change and global ecological degradation, companies increasingly face public pressure to avoid activities which make these problems worse. Many have responded by adopting practices which are legitimately more eco-friendly and beneficial. Yet others, aided by the marketing and public relations industries, have responded in a more cynical way — with campaigns designed to make companies appear more eco-friendly even as they go on degrading the environment. The practice has come to be known as “greenwashing”, and poses a deep challenge for individuals who may seek to reward or punish companies for their good or bad practices. Which companies are truly green? Which are only pretending? How can better data science help with this problem? To find out, Mark Buchanan spoke with data scientist Senna Ma, a technical mentor on the project. “Greenwashing normally goes unnoticed,” she says. “It happens when many companies mislead consumers about how green their product actually is. For instance, on social media they might portray their company as very green, even if they aren’t actually meeting their stated commitments on the environment, social and governance issues.” Other companies may create apparently eco-friendly and green products and sell them for higher fees because people are willing to pay. Yet the companies might not be offering what people expect. “The problem,” says Ma, “is that there’s actually very little information available on whether a company is greenwashing or not.” During the DSSG summer school, Ma and other data scientists attacked this issue with several partners: the Algorithmic Transparency Institute, and two climate researchers, Jeffrey Supron and John Cook, who had worked on a project last year. Mark Buchanan: What is the ultimate goal of this work? Senna Ma: “We’re working on building a way to detect whether or not social media posts, from different companies within the fossil fuel industry, are green, non green, or something else. They could be ambiguous. We hope that by classifying many social media posts we can start to generate a data set which can then be used in, collaboration with other metrics, to really identify whether companies are greenwashing or not.” Mark Buchanan: The algorithms you are using must be trained with some examples of things that are obviously greenwashing. Once you learn to identify these examples, you can then look for others that share the same characteristics? Senna Ma: “Yes. Let me talk about the project approach. Our partners have created a taxonomy, and they have many research assistants who go through lots of social media posts and label them as to whether or not they are green, non green, ambiguous, etc.” Mark Buchanan: These are real real human beings, right? Senna Ma: “Yes. And you can imagine it took a lot of time even to just generate this large data set. From this data we have started to create classifiers for process text. But in future we also plan to move on to images and videos.” Mark Buchanan: How might this eventually be used, once it’s all polished and finished? Senna Ma: “Anyone browsing through advertisements on the web would be able to be alerted that someone has flagged this or that advertisement as being greenwashing. That would be the gold standard.” Or, she suggests, such work might ultimately lead to browser plug in that would automatically tell someone encountering an advertisement online how green that company actually is. Reaching this stage will no doubt take much further work, both by data scientists and by individuals helping to classify advertisements and other corporate messages according to their eco-friendly authenticity. In the long run, such efforts could help restore the ability of people to use the power of purchase to push companies toward more eco-friendly actions.