ESG and AI

In an era marked by unprecedented technological advancement and growing concerns about the planet’s sustainability, the intersection of artificial intelligence (AI) and Environmental, Social, and Governance (ESG) principles has emerged as a powerful catalyst for change. AI’s transformative capabilities have the potential to revolutionize how businesses and society address the critical challenges of our time, from mitigating climate change and fostering social equity to enhancing corporate responsibility. As we stand at the crossroads of innovation and ethics, this article delves into the dynamic relationship between AI and ESG, exploring how these two forces are shaping our future and driving a paradigm shift towards a more sustainable and socially responsible world. 

When AI first hit the virtual shelves, it seemed like a miracle cure. It could draft your emails, comb through long documents, improve business processes and more. If given the prompt “Can you draft an opening paragraph for an article on AI and ESG?” it could even write the paragraph above this one. But while AI has huge potential, it is still just a tool.  

A new industry of ESG-oriented AI tools has grown to support rising demands from financial institutions. As investors increasingly seek to align their values to their investments, asset managers are looking to technology to help through screening funds, portfolios, companies, or investment opportunities. 

 To do this,  AI combs through a large amount of data including nonfinancial reports to gauge a firm’s ESG risks, values alignment, environmental impact, or exposure to socially or environmentally sensitive issues, thus allowing investment managers to divest from or engage with these companies.  

In Islamic finance, this means removing industries that are haram, or not allowed, such as alcohol, pornography, or pork products. In ESG or impact investing, this may mean removing firms in industries like oil and gas or coal. There is a larger debate about whether to engage or divest from these firms, but that is a much larger conversation and two separate discussions pieces (see the GEFI and UKIFC websites for more).  

Beyond investment screening, AI can also be trained to analyse any sort of nonfinancial report, ranging from quarterly earnings to environmental reports. This can quickly process the values and attitudes of firms, which can help investors and consumers decide how to best use their spending power.  

But, as the saying goes, garbage in, garbage out. AI can only read what is published, meaning that if a firm is misrepresenting itself, AI cannot always detect itIf a company is greenwashing and presenting itself as significantly more environmentally friendly than it actually is, this may go undetected.  

AI can be an incredibly useful tool in analysing reports published by firms. During my PhD, I used the coding language R (with some clever codes written by someone else) to examine the readability of ESG reports. As AI can ‘learn’, it can essentially train itself to be more effective at executing tasks and analyses such as these. This could include performing Sentiment Analysis on the tone of CEO forewords to quarterly earnings reports. If the CEO uses defeatist language or expresses negative sentiments towards ESG issues, this can be used to guide investment decisions. 

While AI can analyze large documents with programs like Sentiment Analysis, it is important to remember that awareness creates opportunity. If firms are aware that AI models will be analysing their reports, those reports will over time be written with this analysis in mind, perhaps blunting its effectiveness. 

While AI can be used for environmentally beneficial purposes, the impact of running complex AI is hard on the environment. In a recent study by University of Massachusetts Amherst, the training process for a common large AI model produced 283948 kg  of CO2, the equivalent of 300 round trip flights from San Francisco to New York. With AI riding a boom, the environmental impact is massive. Furthermore, in order to build these massive processing facilities, firms need minerals and metals. There is continued risk that these are sourced from conflict regions, leading to not only environmental degradation, but human rights abuses and mass exploitation as well.  

Beyond the environmental concerns, bias in AI training can lead to unexpected negative real-world implications. For example, the headlining 2018 paper Gender Shades by Dr Joy Buolamwini and Dr Timnit Gebru found that AI misidentified faces at  an error rate for women of color just shy of 35% while the error rate for white men was below 1%. The dataset the AI trained on was hugely biased toward lighter skin and male-presenting faces. 

 Also in 2018, Amazon’s hiring AI came under fire for discriminating against female applicants, as it was trained off the existing datasets of ‘desirable’ qualities in the tech sector, which is predominantly male. This taught the AI to downgrade resumes with any references to women, such as graduating from an all-women’s University or being involved in a ‘women’s’ club. Considering the social element of ESG, this poses a problem if we’re trying to use technology to avoid human bias, but have built human bias into technology. 

The general overuse of AI in situations such as these raises important governance concerns. If important decisions are delegated to inscrutable processes, then corporate accountability for the negative consequences of those decisions could be stymied. 

Even after all this, I, for one, am still an optimist about AI. Any new technology experiences a period of growth and development, often with quite a few speedbumps along the way. With AI being an undisputed part of our collective future, it is best that we recognize these issues early and work through them with openness, awareness, and consideration.  

So what is to be done? Three things; first, utilize regulation to ensure transparency and accountability in firm reporting so that ESG AI will have accurate and high-quality data to draw from; second, be conscious of the environmental impact and creative with solutions, such as using renewables; third, conduct rigorous research into the human element of AI and ensure that there is representation in the organizations designing AI. We all see the world through the lenses of our own experience, so bringing in diversity of experiences to built the technology will help ensure that it is designed with more than one lens.  

AI can be an incredible tool to further the important work being done in the global economy to transition to net zero, bring human rights abuses to light, and make our world an overall better place.