Tommaso Ciarli (p), Önder Nomaler, Bart Verspagen

Eleni Kalantzi

This study investigates the intersection of Artificial Intelligence (AI) and green technologies using scientometrics and generative AI (ChatGPT). By analyzing approximately 43,000 technological trajectories derived from European Patent Office data, we map current trends and project future innovation pathways. Focusing on 950 trajectories explicitly referencing AI/machine learning, a sample of 32 was analyzed in-depth using ChatGPT. The AI summarized patent information, predicted future directions, and identified related technologies. This process revealed key paradigms like "Automation and Control Systems" and "Environmental Management and Sustainability" characterizing past innovations. Projected future paradigms include "Integration with IoT and Smart Systems" and "Enhanced AI and Machine Learning Integration." A tripartite network visualization highlights clusters around areas like mobility, wind turbines, and environmental monitoring. While the pilot study demonstrates the potential of combining scientometrics and generative AI, limitations include the small sample size and replicability challenges. Ongoing work expands the analysis to the full dataset, explores reproducible AI models like Llama, and compares findings with established NLP methods for robust validation. This research ultimately aims to provide a comprehensive understanding of the evolving landscape of AI-driven green technologies.

Eleni Kalantzi (p), Christina Kakderi

Emerald Dilworth

Since 2019, the European Commission (EC) has placed transition to green forms of production and consumption and digital transformation at the core of its strategic vision for growth. This commitment is evident in major policy documents such as the European Green Deal, the Digital Strategy, and the New Industrial Strategy for Europe, which set the foundation for the EU’s twin transition. As the EU continues to prioritize green and digital transition, it is important to understand how the two policy streams engage with inequality dimensions that may emerge or be exacerbated due to the twin transition. This study aims to analyse the policy discourse surrounding the green, digital, and twin transitions in the European Union (EU) by applying topic modelling to a dataset of 100 EU policy documents. The main objectives are to identify key thematic areas, assess their alignment with EU strategies and classify policies into green, digital, or twin categories. The dataset consists exclusively of official EU policy documents retrieved from EUR-Lex, the EU’s legal database. These include communications, white papers, regulations, directives, council conclusions, proposals and reports, ensuring a comprehensive and authoritative representation of EU policy priorities. The methodology employs Natural Language Processing (NLP) techniques for text preprocessing, including tokenization, lemmatization, stopword removal, and Term Frequency-Inverse Document Frequency (TF-IDF) filtering to enhance data quality. We perform topic modelling to uncover the dominant themes of each document, and specifically Latent Dirichlet Allocation (LDA) and BERTopic, using both statistical modelling and AI-driven embeddings to improve interpretability. Extracted topics are then classified into green, digital or twin using machine learning-based classification methods. As a final step, validation techniques such as coherence score analysis, manual review and supervised classification models are applied to ensure the accuracy and reliability of results. The paper manages to uncover underlying themes, priorities, and gaps in policy discussions, revealing how policies attempt to mitigate or manage different types of inequalities related to green and digital transformations.

Emerald Dilworth (p), Emmanouil Tranos, Daniel Lawson

Kateryna Tkach

When individuals and companies struggle to keep up with new technologies, there can be a wide range of barriers stifling their access, and in turn, many consequences from failing to keep up with the cutting edge.
We look to firms involved in green and digital technologies in the UK to better understand how and where twin transition technologies are developing.

Traditionally researchers have relied on surveys, patent data, and place attribute-based data to investigate such problems. Classification systems such as NACE and SIC are often used, but are not updated frequently enough to classify new industries and technologies well.
Alternative data labelling often hides behind a paywall and a lack of transparency on the labelling procedure.
The temporal lag, poor labelling, or paywall of these data poses a challenge to effective policy design which can drive and facilitate the twin transition.

To address this research gap, we use free to access, monthly updated data at the website level to identify and classify firms involved with digital and/or green technologies in the UK.
We utilise the ``digital breadcrumbs'' created simply from the use and adoption of the Web.
From vast quantities of web-crawl data obtained from the Common Crawl, we utilise the webtext to inform if a website has an interest in green and/or digital technologies, by fine-tuning a transformer based language model to label websites, specifically for the task where interested in twin transition technologies.

Analysing the period from 2014 to 2024 allows us to observe temporal changes in twin transition technologies and examine the influence of spatial factors and urban agglomeration on the distribution and evolution of twin transition industries over time.

Michela Borghesi, Alberto Marzucchi, Ugo Rizzo, Kateryna Tkach (p)

Tommaso Ciarli

The digital and green transition, also referred to as “twin transition”, is at the core of current EU policy frameworks and academic discussions. Although there is a steadily growing stream of research on effects of this intertwined transformation, evidence on its implications for local economies is still emerging. While previous studies have focused primarily on the demand-side effects of this transition, the supply-side impact, particularly at more granular geographical levels, has received relatively less attention. This paper aims to contribute to this stream of research by examining the role of digital and green skill supply for industrial dynamics in Italy. Using the data from the mandatory reports submitted by Italian universities and firm entry and exit rates at the provincial level, we investigate the impact of digital and green skill supply for local industrial dynamics. Our dataset covers 106 Italian provinces (NUTS 3 level) from 2018 to 2023. Our findings reveal that the supply of digital and green skills from top graduate programs is a positive and significant factor for firms’ entry rate. Interestingly, green skills remain significant even when controlling for the intangible assets available across provinces, whereas digital skills’ effect seems to vanish for the entry rate. Moreover, we find that digital skill supply is negatively associated with the firm exit rate, while for green skills no significant relationship was identified. These results highlight the distinct characteristics of digital and green domains of the twin transition as well as their role for the local industrial dynamics.