By Neil Sahota 5 minute Read
In 2016, Tide launched Purclean, a new brand of detergent that claimed it was 100% plant-based. However, the National Advertising Division of BBB National Programs analyzed the claim four years later and found that Purclean was only 75% plant-based. While not great, 25% non-plant composition doesn’t sound too bad until you learn that some of the materials are petroleum-based. This was completely counter to Tide’s marketing message, and extremely misleading for consumers.
And it’s a classic example of greenwashing, which by definition refers to misleading communication about a company’s environmental practices and impact so as to present an environmentally responsible public image. In a time when marketers have roughly three seconds to grab someone’s attention, it’s a lot easier to spin the truth, especially when it comes to lauding the efforts of sustainability and eco-friendly endeavors. While there are companies committed to making a real difference for people and the planet (like Patagonia or Cree), there are many enterprises that espouse being green more so in marketing than actual practice. But how do we differentiate between greenwashing spin and the true green initiatives when it is incredibly difficult to hold companies accountable for their actions? Thankfully, we have a friend in artificial intelligence.
Meet ClimateBert, an AI tool that deconstructs corporate statements, annual reports, claims, and other materials to assess climate-related disclosures and measure actual performance. It was created by the Task Force on Climate-Related Financial Disclosures (TCFD), which provides a framework for public organizations to more effectively disclose climate-related performance. Because extracting salient information from companies on their climate disclosures is complex and time consuming, TCFD turned to natural language processing and existing deep neural networks for help. The sheer volume of data, often using subtle words, presents a major challenge to analyze in a timely fashion. Thanks to AI tools like ClimateBert, we can now shrink weeks of analysis into just days.
What did ClimateBert discover? Regrettably, after assessing more than 800 companies, ClimateBert has determined that corporations are talking a good game, but actual performance is lacking. Why? In TCFD’s assessment, there are three major contributing factors. First, greenwashing has largely escaped scrutiny so far, so there’s no incentive for companies to change. Second, the Paris accords have, ironically, let companies be more “selective” in what they want to disclose to limit brand risk. Third, with the exception of France, the reporting of corporate climate is a voluntary disclosure, enabling companies a lot of latitude on what they would like to share. That’s why TCFD has been pushing to make reporting standardized and mandatory.
Other organizations are also tapping into the power of AI to discover greenwashing. For example, Ping An, an insurance and finance company located in China, is leveraging its Digital Economic Research Center to use AI to assess corporate climate disclosure and detect greenwashing. Using natural language processing algorithms, the Digital Economic Research Center developed AI-driven indicators to determine climate risk exposure that was more granular than traditional environmental, social, and corporate governance (ESG) metrics. In effect, this AI found a more efficient way to determine if an enterprise was truly being eco-friendly or just greenwashing. Moreover, the AI can dynamically assess, in real time, the actual sustainability practices of a company as it keeps sharing more information.
While these examples sound promising in holding companies accountable to their environmental promises, challenges still remain. Our first problem is meaningful, robust data, which provides the fuel for any AI system to learn what greenwashing looks like. We need good data to train our AI systems as well as to give the machine something to analyze and review. While corporate social responsibility goals have been around for a couple of decades, collecting data on performance has lagged in part because of nebulous or subjective metrics. However, thanks to other emerging technology like IoT sensors (to collect ESG data) and blockchain (to track transactions), we have the infrastructure to collect more data, particularly for machine consumption. By measuring real-time energy usage, transportation routes, manufacturing waste, and so forth, we have more quantifiable ways to track corporations’ environmental performance without relying purely on what they say.
The second problem is applying macro benefits to micro solutions. It is not sufficient or accurate to evaluate corporations’ environmental progress on popular initiatives like tree planting. Companies like Microsoft, Alibaba, American Express, and others are all engaged in programs to plant millions of trees, which sounds like a great idea until you start to consider how much impact it really has. The average mature tree can offset about 48 pounds of carbon per year, but most companies don’t factor in how much time it takes for a tree to grow. Moreover, the species of a tree also dictates how much carbon sequestration occurs. A mature silver maple tree can offset around 500 pounds of carbon per year, while palm trees average around 15 pounds per year. Companies need to understand how many trees, which type of trees, the location of trees, and so forth to accurately count carbon sequestration. This suddenly becomes a more arduous and taxing process that costs enterprises more money, resources, and time, which tends to de-incentivize them from accurately measuring the impact of their so-called eco-friendly initiatives.
Thankfully, AI technology is ideally suited to handling these tasks. With tools like Pachama and ML CO2 Impact, we have AI to assist organizations in accurately measuring and communicating their carbon impacts and offsets at a more granular level. In addition, organizations like Planet Home are using machine learning to develop personalized calculators to measure individual or organizational sustainable behavior to simplify data collection, measurement, and reporting. Moreover, they are helping people identify small steps that they are willing to take to be more sustainable, attempting to go beyond reactive measurement to proactive behavior.
This is the real value we can tap into through AI. Through greenwashing detection, AI helps us build truth and trust in corporate communication. As we shift to a fully integrated, sustainable corporate culture, AI can help organizations find more environmentally friendly opportunities to improve their carbon footprint. Ultimately, using AI to hold companies accountable for their environmental impact and to help them find ways to actually be green will lead to a more sustainable world for everyone.
Neil Sahota is the author of Own the A.I. Revolution: Unlock Your Artificial Intelligence Strategy to Disrupt Your Competition and works with the United Nations on the AI for Good Global Summit initiative. Sahota is also an IBM Master Inventor, former leader of the IBM Watson Group, and professor at the University of California, Irvine.