In brief
AI checks corporate sustainability
Researchers have used a large language model to analyse the annual reports of 600 European companies.
Environmentally friendly: A photovoltaic system on the roof of Soyana, the Swiss vegetarian foodstuff manufacturer. | Photo: Christian Beutler / Keystone
In Europe, companies have to provide information on their sustainability status. They can do this, for example, in their annual reports. Stefan Feuerriegel is an SNSF-funded researcher working at Munich University who has being employing an open large-language model to extract this information from the reports of the 600 biggest companies for the past ten years. This meant working with a total of almost one million data records. “Sifting out this information by hand would have been impossible in sheer terms of the time needed”, he says.
Analysing these figures showed that much has been achieved in terms of environmental friendliness and transparency. There’s been less progress in terms of social sustainability goals, albeit with one exception: closing the gender pay gap. His complete programming code and his datasets are now on open access.
