Camilla Lenzi (p), Roberta Capello, Elisa Panzera

Akop Akopyan

This paper studies the redistributive consequences of digitalisation and highlights the role of urbanization economies as multipliers of wage inequalities. Elaborating on the difference between the new digital technologies and the past waves of computerisation and robotisation in terms of labour market outcomes, the paper argues that digitalisation is expected to generate both a productivity and a reinstatement effect for both high- and low-skilled occupations. The balance between the productivity and the reinstatement effects, however, varies across occupations. For low-skilled workers the reinstatement effect prevails over the productivity effect, while the opposite occurs for high-skilled workers, thus amplifying wage inequalities further. These unbalanced effects are especially prominent in urban areas, which therefore experience a sharper increase of wage inequalities because of the significant diffusion of new digital technologies in highly diversified labour markets. These expectations are tested in an analysis on Italian cities (i.e., NUTS3 regions) in the period 2012-2019.

Simona Ciappei (p), Annalisa Caloffi

Camilla Lenzi

Literature has found that knowledge-intensive business service organizations (KIBS) play a crucial role in driving technological innovation and overcoming barriers to the adoption of emerging technologies. This paper delves into the contemporary digital landscape and investigates the firm-level and territorial-level determinants that affect the engagement of these organizations in the provision of new digital technologies and related services, and how these determinants vary by type of technology. The research considers a subset of technological KIBS (t-KIBS) with a NACE code associated with the provision of ICT technologies. Looking at Italy, an original dataset was built and enriched with data extracted from t-KIBS websites and collected through web-scraping procedures. A keyword-based approach allowed for the identification of the t-KIBS that could be involved in the provision of new digital technologies. A probit model with sample selection was then applied to find out which are the firm-level and territorial-level determinants that affect the probability to provide digital technologies. The results suggest that overall, the involvement in the provision of new digital technologies is strongly affected by the co-location with manufacturing companies that demand these technologies. However, this finding does not hold for all the technologies as some of them rely more on urban advantages. This suggests that in new digital technologies the spatial dimension continues to be important, but in different ways depending on the specific technological context.

Genghao Zhang (p)

Simona Ciappei

Productive economic actors tend to adopt artificial intelligence (AI) technologies selectively since AI improve companies' or industrial productivity to some extent. Limited studies investigate local neighborhood effects on spatial concentration of AI adoption. My research examines how places’ characteristics heterogeneously influence AI location quotients (i.e., the proxy of relative concentration levels), using data of AI online job vacancies between 2016-2021. Our two-level (random) regression model finds significant compositional and contextual effects over time at the LSOAs and TTWAs level, respectively, in the UK. For one thing, intensive AI online hiring activities hinge on places’ characteristics, for instance, urban infrastructure, intellectual resources, social capital, and productivity. For another, more developed regions (i.e., larger travel to work areas) manifest labor demands for AI skills to a relatively greater extent on average across the UK. At the early stage of AI adoption, new yearly AI online jobs become more and more spatially concentrated in specific local areas in each year, indicating potential inequality issues of urban and regional economic development. These research findings should provide local policy makers with place-based and people-based policy recommendations to cultivate places’ characteristics for widespread AI adoption.

Keywords: Compositional effects; Contextual effects; Spatial Heterogeneous; AI online jobs; AI divide

Emmanouil Tranos (p)

Genghao Zhang

This paper maps and models the participation in the digital economy and its evolution in the UK over space and time. Most of the existing economic geography literature which dealt with the spatiality of the internet employed supply-side measures, such as infrastructural capacity, in order to understand the geography of the digital economy and its potential spatial economic effects. Useful as these approaches might have been, they cannot capture the micro-processes and the characteristics of the individual online behaviour. Using large volumes of archived and geolocated web content, this paper models the diffusion of web technologies over space and time in the UK. Instead of using metrics capturing the passive en- gagement with digital technologies, for instance internet subscription metrics, this paper targets the active engagement with the digital as reflected in website creation.

Importantly, the data and geolocation strategy allow to capture these processes at small spatial scales. This level of granularity differentiate this paper with previous approaches in the literature, which were only able to capture technological adoption at more coarse level. Thus, this paper tests how well established theoretical approaches regarding the diffusion of new technologies are still applicable when the focus is on local scales. Although we know that spatial contagion and urban hierarchies are key drivers of technological diffusion, such theoretical concepts have not been tested at small geographical scales. Apart from an empirical interest, understanding such processes at small scales can support designing strategies for the adoption of new technologies and the development of relevant infrastructure.

Akop Akopyan (p), Ruben Gevorgyan

Emmanouil Tranos

The year 2023 marked a significant shift in the field of Artificial Intelligence (AI), with the introduction and growing popularity of generative models such as Chat GPT, Dall-E, Google Bard, Midjourney, and others. This evolution ushered in a new era where businesses rapidly adopted AI technologies into their products, and prompt engineering became the foundation for numerous innovative solutions. While high-tech companies swiftly integrated these new technologies, public sentiment towards AI remained complex and multifaceted. As with any unfamiliar concept, people’s reactions to these revolutionary products varied widely, ranging from uncertainty and fear to excitement and anticipation. These emotions significantly influence public behavior, making it crucial to understand public perceptions of AI. Understanding these perceptions is not only pivotal in shedding light on how generative AI is shaping society but also influences trust, adoption, decision-making, and the ethical development of AI technologies. Furthermore, it highlights areas for education and awareness to align AI advancements with public interest and societal values. This paper seeks to answer the central question: ‘How has public sentiment towards AI evolved over the last two years?’. Employing multidimensional sentiment analysis techniques, this study analyzes a corpus of 639 most read world news articles, collected via the G-News API, published before and after the Generative AI Boom in 2023. In this study bart-large-mnli model of Facebook is used to estimate different sentiment scores of the articles using zero-shot classification techniques. The analysis captures various dimensions of public perceptions, such as fear and excitement, job displacement alerts, the tone of voice of the articles and overall trust in AI. These sentiment scores are then analyzed as time series, considering major AI events as influencing factors. The results reveal periods of uncertainty about AI, alongside clear trends of positive or negative sentiment shifts, and an overall pattern of the public sentiment formation. This paper introduces a novel methodology for studying public perceptions using textual data, contributes to ethical AI discussions, and provides insights into future trends.