AI Tools Uncover Hidden Corporate Trends in Responsibility, Innovation, and Blockchain

Using advanced machine learning and AI-driven text analysis, a Finnish researcher reveals how companies frame innovation and responsibility—and why it may not always reflect reality.

​​​​​​​Research: Essays on Corporate Textual Disclosure. ​​​​​​​Image Credit: NicoElNino / Shutterstock

Machine learning methods enable the efficient mining of large amounts of data from corporate reports. In her doctoral dissertation at the University of Vaasa, Finland, Essi Nousiainen presents how her new machine learning-based methods can be used to examine corporate reporting related to responsibility, innovation, and blockchain.

Corporate accounting reports contain extensive data on various themes, and machine learning methods and AI-based language models offer entirely new possibilities for exploring this data. In her doctoral dissertation in Accounting, Essi Nousiainen illustrates what information can be gathered about corporate responsibility, innovation, and blockchain-related trends using machine learning and AI-based text analysis. The analyses are based on measurement methods developed by Nousiainen.

The text analyses revealed, for instance, that companies publicly seeking buyers report more on responsibility than their peers. However, the actual responsibility actions of these companies did not differ from those of the comparison group. This suggests an attempt to appear more responsible in sales situations and highlights the need for responsibility reporting regulation, says Nousiainen, who will defend her dissertation at the University of Vaasa on 4 April.

She adds that the analyses also showed that companies are addressing topics related to cryptocurrencies more cautiously than before, while the trend was the opposite for other blockchain-related topics.

New metrics help with competitor and industry analyses

Nousiainen's dissertation develops new machine learning and text-based metrics for measuring corporate innovation and responsibility from accounting reports.

With the innovation metric, report topics can be mined and compared, allowing the level of corporate innovation to be identified without just examining patents. The responsibility metric, on the other hand, measures how much companies report on their responsibility based on keywords and contexts.

The dissertation also introduces a research method for analysing corporate blockchain and cryptocurrency reporting, where existing machine learning-based analysis techniques have been combined in a new way.

Nousiainen summarises that companies, researchers, and anyone interested in financial statement data can use these different metrics and methods. For example, companies can use the methods in competitor and industry analyses, during mergers and acquisitions, and when seeking business partners.

Nousiainen used 10-K and S-1 reports—that is, annual reports and listing particulars—from U.S. companies as her research material. Her research methods included Latent Dirichlet Allocation (LDA), sentiment analysis, and statistical modeling.