AI starts to show promise as tool to sift mountain of sustainability research

Technology is starting to challenge the supremacy of human judgment in measuring the impact of business school academics’ research and teaching on sustainability.

The value of academic papers is often measured by the perceived prestige of the journal the work is published in. But artificial intelligence models are starting to be used to unearth and flag up a broader range of papers among the many thousands of articles published every year.

The leading commercial providers of academic publishing statistics — Web of Science and Scopus — as well as researchers at St Joseph’s University and elsewhere, have, in recent years, developed ways to automatically determine the relevance of journal articles to the UN Sustainable Development Goals.

Similarly, Wilfred Mijnhardt, policy director at the Rotterdam School of Management at Erasmus University, has created a series of tools based on machine learning to score academic articles for their relationship to the SDGs.

This exercise shows that AI could be useful in finding papers that would otherwise be overlooked, although it still makes basic errors, such as flagging papers that include the word “climate” even when unrelated to sustainability.

“Machine learning is like a baby,” says Mijnhardt. “But it’s growing much faster than humans.”

In co-operation with the FT, he applied his analysis to two sets of publications: the FT50 list of top journals and the longer Chartered Association of Business Schools list of the top A and A* journals. Both lists are drawn up on the advice of top academics.

There is debate about how wide to cast the net among academic journals. Some say a broad range should be picked in order to include emerging, interdisciplinary or overlooked work, as well as research from outside the western, English-speaking world.

Machine learning is like a baby. But it’s growing much faster than humans

An alternative approach is to focus on the most prestigious and rigorous journals, which command the greatest respect within business schools, pushing this elite group to pay more attention to new trends and insights on sustainability.

Using the smaller FT50 list of journals as a proxy for this second approach, and focusing on articles published since 2020 relevant to the climate change SDG, Mijnhardt’s software picked: Corporate Commitment to Climate Change: The Effect of Eco-Innovation and Climate Governance, by authors at Portsmouth, Newcastle and Henley business schools.

That article found that quoted UK companies with environmental committees, climate incentives and sustainability reporting have a greater commitment to climate change and are able to achieve greater “eco-innovation”.

Another top-scoring article explored how to balance sustainability considerations with urgent but polluting humanitarian interventions, such as producing syringes for vaccination and plastic bottles of water.

But the software also flagged up irrelevant articles, including one about entrepreneurship that contained the phrase “environmental changes — be they technological, regulatory, demographic, sociocultural, or otherwise”.

Applying Mijnhardt’s software to the longer CABS list of academic journals produced additional papers, such as articles on the effects of climate change on hospital admissions in Inuit communities, and on maternal healthcare in rural Bangladesh. Some wrongly identified the authors as affiliated to a business school, or jumped on terms such as “socio-economic climate”.

Another, simpler approach to evaluating the impact of business school academics’ research, is to count the number of downloads from a popular database. The free-to-use Social Science Research Network hosts uploads of papers by researchers, before they are published in academic journals.

Working with the FT, SSRN identified the most popular business-school research papers downloaded by users in positions of influence, including commercial and central banks, regulators, and local and national government agencies.

Top of the ranking was a paper on the “Issuance and Design of Sustainability-Linked Loans”, by academics at Harvard and the University of Texas at Dallas. It showed that the pricing of such loans was usually based on weak targets that did not link to ESG conditions most relevant to the borrower so the loans were unlikely to result in sustainability improvements.

The next most popular was “Can Investors Save the Planet?” from two UK schools, which concluded that the strategies endorsed by the Net Zero Asset Managers Initiative that are most likely to be adopted are also those least likely to contribute meaningfully to addressing climate change.

These exercises show the potential, but also the limitations, of technology to help in the analysis and selection of academic research.

Another potential application of technology is not to research but to teaching — by tracking the extent and content of courses provided by business schools, and the specialisms of those who teach them.

Coursalytics, a digital executive education consultancy, surveyed 120 professors, examined 14,000 articles, and analysed the syllabuses of 168 ESG-related executive and graduate-level programmes taught at business schools and universities since 2020, using web scraping and scrutiny.

It concluded that they mostly focused on corporate governance, responsible business practices, and the green agenda. Participants said that environmental sustainability, corporate governance and social impact had the greatest priority over the next two years.

“In general, ESG education does not capture the business demand in full so far,” says Ilya Breyman, Coursalytics’ chief executive.

Machine learning, and more basic digital support and metrics, offer growing potential to benchmark and mobilise the sector. But, for now, human qualitative judgment remains essential for assessing the most meaningful and valuable research and teaching.