This dissertation, presented as a collection of three papers, explores the intersection of artificial intelligence (AI) technologies and sustainable investing. It examines the state of the art, potential benefits and limitations, and the ethical and practical challenges arising from the integration of AI-driven systems into areas traditionally governed by human judgement in sustainability rating and investment paradigms. This relationship is investigated across three key domains of the sustainable investment ecosystem: investment processes, ESG (environmental, social, governance) rating agencies, and issuers (i.e., companies receiving ESG ratings). For sustainable investing to fulfil its promise of driving a long-term shift towards sustainability, ESG criteria must be measured alongside financial processes. Reliable ESG data has thus become more critical than ever in financial markets, fuelling the expansion of ESG rating providers in the market. These providers offer assessments, rankings, and insights into companies’ sustainability risks, opportunities, and performance. Today, trillions of dollars are invested based on ESG ratings, yet concerns about their transparency, reliability, comparability, and consistency persist, raising doubts about their validity as sustainability measurement tools. On the supply side of ESG ratings, companies receiving them often prioritise compliance with rating criteria over substantive sustainability improvements, reinforcing concerns that ESG ratings may be adhering to financial market expectations rather than driving real change in corporate behaviour. Recent advances in AI have introduced promising solutions for investors to process vast amounts of ESG data efficiently, particularly as sustainability information is often found unstructured, unstandardised, and qualitative forms. Some ESG rating providers have already integrated AI-driven assessment mechanisms, contributing to the rapid expansion of the ESG data and analytics sector. However, the lack of standardisation, potential biases in AI models, and ethical risks raise concerns about the true impact of AI-driven measurement systems on sustainability assessments. Given these challenges, both sustainable investment research and practice are at a crossroads in determining how AI can enhance investment strategies. Despite its potential, the role AI plays in sustainable investments and corporate sustainability assessment remains unclear. This dissertation aims to determine whether and how AI helps preserve—or potentially undermines—the integrity of sustainable investments. The research problem is examined in the empirical contexts of ESG rating providers and listed European companies. Given the exploratory and emerging nature of the topic, a qualitative approach based on the Gioia method (Gioia et al., 2013) and abductive thematic analysis is adopted for data analysis. The first paper systematically reviews the literature on AI applications in sustainable investing, identifying key areas where AI is being used and assessing its benefits and limitations. Mapping these findings onto a conceptual investment process framework, the paper provides a common ground for academics and practitioners, fostering knowledge sharing. The results reveal a significant shift from traditional, backward-looking, manual processes to advanced, automated, forward-looking models that leverage machine learning and natural language processing to analyse ESG disclosures more efficiently and integrate alternative data sources. The second paper, based on interviews with ESG rating providers, examines the interplay between AI and human analysts in sustainability measurement processes. The findings identify four key dimensions that define their respective roles: (i) AI’s catalyst role for measurement; (ii) cognitive interpretation of measurement by humans; (iii) deterministic drift of technology; (iv) contingencies as exogeneous factors. While AI’s data processing capabilities are transformative, the study points out the ongoing need for human cognitive and critical thinking skills to ensure nuanced interpretation of sustainability in measurement mechanisms. The findings emphasises the trade-off between quantity and quality in ESG assessments, where an overreliance on AI without human oversight may lead to ethical issues. The third paper shifts the focus to the issuer perspective, investigating why ESG ratings remain widely adopted despite persistent criticisms. Drawing on semi-structured interviews with listed European issuers, the findings suggest that ESG ratings are becoming institutionalised, driven by financial sector pressures and consequently the adherence from issuers, both symbolically and substantively. This institutionalisation creates both isomorphism costs and opportunities, with normative and cultural pressures shaping how companies engage with ESG ratings. The paper argues that while ESG ratings enhance transparency in the market and foster relationships with stakeholders, they also impose compliance costs and legitimacy pressures on issuers, which, in turn, may lead to the de-institutionalisation of measurement practices. Finally, AI is expected to play an increasingly central role in sustainable finance, particularly as non-financial disclosure regulations expand and stakeholder expectations intensify. However, striking a balance between AI’s precision and scalability and the ethical challenges it poses remains critical. Future research must investigate how AI integration is reshaping sustainable investment decisions and capital allocations, and how these evolving systems can be understood, regulated, and trusted in institutional settings. Addressing these challenges requires coordinated action among academics, investors, technology providers, and policymakers to establish clear governance frameworks, reliable data standards, and ethical oversight, as essential to ensuring that AI technologies are deployed responsibly within sustainability paradigms.
La tesi esplora il rapporto tra l’intelligenza artificiale (IA) e gli investimenti sostenibili attraverso tre articoli, analizzando benefici, limiti e criticità etiche nell’integrazione dell’IA nei processi di rating e investimento ESG. L’analisi empirica si concentra su investitori, agenzie di rating ESG e aziende emittenti. Gli investimenti sostenibili necessitano di dati ESG affidabili per garantire decisioni informate. Tuttavia, la trasparenza e l’affidabilità dei rating ESG vengono spesso criticate, sollevando dubbi sulla loro efficacia nel promuovere pratiche sostenibili. Le tecnologie di IA offrono soluzioni per elaborare grandi volumi di dati ESG, migliorando l’analisi predittiva. Tuttavia, la poca trasparenza degli algoritmi nel valutare i temi legati alla sostenibilità e il rischio di bias pongono interrogativi sul loro reale impatto. Il primo articolo esamina la letteratura sull’uso dell’IA negli investimenti sostenibili, evidenziando un passaggio da processi manuali a modelli automatizzati. Il secondo, basato su interviste con esperti di rating ESG, esplora l’interazione tra IA e analisti umani, sottolineando il trade-off tra quantità e qualità nelle valutazioni ESG. Il terzo analizza il punto di vista delle aziende emittenti, evidenziando la crescente istituzionalizzazione dei rating ESG e le pressioni normative e di mercato. I risultati suggeriscono che l’IA sarà sempre più integrata nei processi di sostenibilità finanziaria, ma è necessario un equilibrio tra precisione, scalabilità ed etica. Una regolamentazione chiara e una governance solida sono essenziali per garantire che l’IA rafforzi l’integrità degli investimenti sostenibili, promuovendo decisioni basate su dati affidabili e trasparenti.
The role of artificial intelligence in sustainable investing
KOPAL, ZEYNEP HAZAL
2024/2025
Abstract
This dissertation, presented as a collection of three papers, explores the intersection of artificial intelligence (AI) technologies and sustainable investing. It examines the state of the art, potential benefits and limitations, and the ethical and practical challenges arising from the integration of AI-driven systems into areas traditionally governed by human judgement in sustainability rating and investment paradigms. This relationship is investigated across three key domains of the sustainable investment ecosystem: investment processes, ESG (environmental, social, governance) rating agencies, and issuers (i.e., companies receiving ESG ratings). For sustainable investing to fulfil its promise of driving a long-term shift towards sustainability, ESG criteria must be measured alongside financial processes. Reliable ESG data has thus become more critical than ever in financial markets, fuelling the expansion of ESG rating providers in the market. These providers offer assessments, rankings, and insights into companies’ sustainability risks, opportunities, and performance. Today, trillions of dollars are invested based on ESG ratings, yet concerns about their transparency, reliability, comparability, and consistency persist, raising doubts about their validity as sustainability measurement tools. On the supply side of ESG ratings, companies receiving them often prioritise compliance with rating criteria over substantive sustainability improvements, reinforcing concerns that ESG ratings may be adhering to financial market expectations rather than driving real change in corporate behaviour. Recent advances in AI have introduced promising solutions for investors to process vast amounts of ESG data efficiently, particularly as sustainability information is often found unstructured, unstandardised, and qualitative forms. Some ESG rating providers have already integrated AI-driven assessment mechanisms, contributing to the rapid expansion of the ESG data and analytics sector. However, the lack of standardisation, potential biases in AI models, and ethical risks raise concerns about the true impact of AI-driven measurement systems on sustainability assessments. Given these challenges, both sustainable investment research and practice are at a crossroads in determining how AI can enhance investment strategies. Despite its potential, the role AI plays in sustainable investments and corporate sustainability assessment remains unclear. This dissertation aims to determine whether and how AI helps preserve—or potentially undermines—the integrity of sustainable investments. The research problem is examined in the empirical contexts of ESG rating providers and listed European companies. Given the exploratory and emerging nature of the topic, a qualitative approach based on the Gioia method (Gioia et al., 2013) and abductive thematic analysis is adopted for data analysis. The first paper systematically reviews the literature on AI applications in sustainable investing, identifying key areas where AI is being used and assessing its benefits and limitations. Mapping these findings onto a conceptual investment process framework, the paper provides a common ground for academics and practitioners, fostering knowledge sharing. The results reveal a significant shift from traditional, backward-looking, manual processes to advanced, automated, forward-looking models that leverage machine learning and natural language processing to analyse ESG disclosures more efficiently and integrate alternative data sources. The second paper, based on interviews with ESG rating providers, examines the interplay between AI and human analysts in sustainability measurement processes. The findings identify four key dimensions that define their respective roles: (i) AI’s catalyst role for measurement; (ii) cognitive interpretation of measurement by humans; (iii) deterministic drift of technology; (iv) contingencies as exogeneous factors. While AI’s data processing capabilities are transformative, the study points out the ongoing need for human cognitive and critical thinking skills to ensure nuanced interpretation of sustainability in measurement mechanisms. The findings emphasises the trade-off between quantity and quality in ESG assessments, where an overreliance on AI without human oversight may lead to ethical issues. The third paper shifts the focus to the issuer perspective, investigating why ESG ratings remain widely adopted despite persistent criticisms. Drawing on semi-structured interviews with listed European issuers, the findings suggest that ESG ratings are becoming institutionalised, driven by financial sector pressures and consequently the adherence from issuers, both symbolically and substantively. This institutionalisation creates both isomorphism costs and opportunities, with normative and cultural pressures shaping how companies engage with ESG ratings. The paper argues that while ESG ratings enhance transparency in the market and foster relationships with stakeholders, they also impose compliance costs and legitimacy pressures on issuers, which, in turn, may lead to the de-institutionalisation of measurement practices. Finally, AI is expected to play an increasingly central role in sustainable finance, particularly as non-financial disclosure regulations expand and stakeholder expectations intensify. However, striking a balance between AI’s precision and scalability and the ethical challenges it poses remains critical. Future research must investigate how AI integration is reshaping sustainable investment decisions and capital allocations, and how these evolving systems can be understood, regulated, and trusted in institutional settings. Addressing these challenges requires coordinated action among academics, investors, technology providers, and policymakers to establish clear governance frameworks, reliable data standards, and ethical oversight, as essential to ensuring that AI technologies are deployed responsibly within sustainability paradigms.| File | Dimensione | Formato | |
|---|---|---|---|
|
36_KOPAL_GIORGINO.pdf
non accessibile
Descrizione: The dissertation "The role of artificial intelligence in sustainable investing" includes the cover letter and the manuscripts of the three papers.
Dimensione
2.31 MB
Formato
Adobe PDF
|
2.31 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/237094