The article in the url below reports what it presents as a breakthrough in the measurement of firms' ESG performance. The article sets up its pitch with breathtaking claims about the connection between ESG metrics and firm performance, while at the same time (and apparently without any irony) acknowledging that we do not yet know how to measure the things that are 'known' to be so crucial:
"It is indisputable that company reputation and performance in the long term is linked to environmental, social and corporate governance performance. … However, quantifying ESG performance on a global scale is a deeply complex problem that was unsolved and unstandardized."
The breathlessness continues:
"For the first time, it is possible to create a meaningful global framework for ESG using artificial intelligence and machine learning under the assumption that a company's ESG perception is a proxy for its ESG performance."
Here's how it works:
"Take a scenario where you are able to analyze vast amounts of textual content from the world's media and other external sources, such as legislation, company releases and regulatory sources, and then measure the perception of an organization. Then cross-reference this perception with specific aspects of ESG that an organization is targeting at a point of time such as conservation or labor practices. Let's call these ESG topics pillars. You can then compare pillars between peers, gather actionable insight and improve the organization's ESG performance, and ultimately affect critical decision making. Using AI, analysts can now search vast amounts of information, from millions of documents from around the globe in virtually any language to understand the ESG reputation of organizations and distill it into a 'reputation score' for different ESG pillars."
Although the stated goal is compatibility and standardization, the author immediately undermines themselves by claiming that an advantage is its customization:
"Relevant ESG pillars can fit into wider industry standards such as SASB or UNSDG. Alternatively, individuals and companies can easily train their own AI models, personalizing their results according to their own interpretation and definition of ESG pillars."
The author would have benefitted from a first year Ph.D. methods course. The A.I. is impressive in the amount of data it can process and allows the resulting measurement to be accurate and consistent (both essential elements of an effective measure), but fails miserably in terms of construct validity. In essence, it is simply a more powerful measurement tool. It does nothing to ensure that the 'right' things are being measured. The article recognizes its own problem when it states that their premise holds, only "under the assumption that a company's ESG perception is a proxy for its ESG performance." That is a massive assumption that I do not think holds at all. What they are doing is accurately capturing what people think firms are doing – not what they are actually doing, or even what they necessarily 'should' be doing. As such, this should not be heralded as a step in the right direction. More likely, it will lead to the punishment of firms that are thought to be doing badly, irrespective of what they are actually doing.
Having a more powerful measurement tool is useless if the thing you are measuring is the 'wrong' thing. Much more important is to do the upfront thinking about what we should be measuring, and then work out how to measure those things more effectively.
Take care
David
David Chandler
© Sage Publications, 2020
Instructor Teaching and Student Study Site: https://study.sagepub.com/chandler5e
Strategic CSR Simulation: http://www.strategiccsrsim.com/
The library of CSR Newsletters are archived at: https://strategiccsr-sage.blogspot.com/
All hail the AI-fueled ESG framework
By Miguel Martinez
December 11, 2020
PR Week
