Abstract
This dissertation examines the differential impacts of artificial intelligence (AI) and automation on labour markets across geographic areas, industrial sectors, and educational attainment levels. Employing a comprehensive literature synthesis methodology, this study analyses contemporary peer-reviewed research to identify patterns of technological displacement and augmentation affecting workers and regions asymmetrically. The findings reveal that AI and automation do not affect all workers or places equally; rather, these technologies tend to benefit high-skilled, AI-complementary workers and advanced sectors whilst raising displacement risks for low-skilled workers in routine occupations and economically disadvantaged areas. Geographic analysis demonstrates that large, high-skill cities capture productivity gains whilst smaller cities and rural areas bear disproportionate vulnerability to automation-induced job losses. Sectoral analysis indicates that routine, manual occupations face displacement whilst cognitive, high-skill work experiences augmentation. Educational attainment emerges as a critical mediating factor, with automation significantly increasing returns to higher education and widening wage gaps. The research concludes that targeted policy interventions around education, reskilling, and place-based support are essential for achieving an inclusive technological transition.
Introduction
The rapid advancement of artificial intelligence and automation technologies represents one of the most significant transformations affecting contemporary labour markets. Since the early 2010s, developments in machine learning, robotics, and cognitive computing have accelerated at an unprecedented pace, fundamentally reshaping the nature of work across virtually every sector of the economy (Brynjolfsson and McAfee, 2014). These technological changes carry profound implications for employment patterns, wage structures, and broader economic inequality, yet their effects are distributed unevenly across different populations and places.
Understanding how AI and automation impact the labour market differently by geographic area, economic sector, and educational attainment has become an urgent priority for policymakers, researchers, and practitioners alike. The stakes are considerable: poorly managed technological transitions risk exacerbating existing inequalities, concentrating economic benefits among already advantaged groups whilst leaving vulnerable populations further behind. Conversely, well-designed interventions could harness these technologies to promote more inclusive prosperity.
The academic significance of this topic extends across multiple disciplines, including economics, geography, sociology, and public policy. Labour economists have long debated the relationship between technological change and employment, with perspectives ranging from techno-optimism regarding job creation to concerns about technological unemployment (Autor, 2015). Geographers and regional scientists have emphasised the spatial dimensions of economic restructuring, documenting how previous waves of automation concentrated negative effects in specific regions whilst benefiting others. Education researchers have examined how skill requirements evolve alongside technological change, with implications for training systems and lifelong learning policies.
Socially and practically, the differential impacts of AI and automation carry significant consequences for individual livelihoods, community wellbeing, and social cohesion. Workers in routine occupations face uncertainty about their employment futures, whilst those with AI-complementary skills command increasing wage premiums. Local labour markets that host concentrations of automatable jobs confront the prospect of economic decline, whilst those with knowledge-intensive industries attract investment and talent. These divergent trajectories risk deepening spatial inequality and undermining social mobility.
This dissertation provides a systematic examination of how AI and automation affect labour market outcomes across the dimensions of geography, sector, and education. By synthesising contemporary research evidence, it aims to identify key patterns and mechanisms underlying differential technological impacts, thereby informing more effective policy responses.
Aim and objectives
The overarching aim of this dissertation is to analyse and critically evaluate the differential impacts of artificial intelligence and automation on labour markets across geographic areas, industrial sectors, and educational attainment levels.
To achieve this aim, the following specific objectives guide the research:
1. To examine and synthesise existing evidence regarding geographic variations in AI and automation impacts, with particular attention to differences between urban and rural areas, large and small cities, and prosperous and disadvantaged regions.
2. To analyse sectoral patterns of technological exposure, distinguishing between industries and occupations subject to displacement effects versus those experiencing augmentation and new task creation.
3. To investigate how educational attainment mediates the relationship between technological change and labour market outcomes, including employment prospects, wage levels, and skill premiums.
4. To evaluate the implications of differential technological impacts for labour market inequality across multiple dimensions, including between-group wage gaps, within-occupation wage dispersion, and spatial economic divergence.
5. To identify policy implications and potential interventions that could promote more inclusive technological transitions across places, sectors, and educational groups.
Methodology
This dissertation employs a literature synthesis methodology to address its research aim and objectives. Literature synthesis represents an established and rigorous approach within the social sciences, enabling researchers to systematically analyse, integrate, and critically evaluate existing research findings to develop comprehensive understanding of complex phenomena (Snyder, 2019). This methodology is particularly appropriate for examining the differential impacts of AI and automation, given the substantial and rapidly growing body of empirical research addressing various dimensions of this topic.
The literature synthesis proceeded through several structured stages. First, relevant peer-reviewed publications were identified through systematic searches of academic databases, with search terms encompassing artificial intelligence, automation, robotics, labour markets, employment, wages, inequality, geographic variation, sectoral impacts, and educational attainment. Sources were selected based on explicit quality criteria, including publication in peer-reviewed journals, methodological rigour, and relevance to the research objectives.
Second, identified sources were systematically reviewed and categorised according to their primary focus: geographic/spatial impacts, sectoral/occupational impacts, education-related impacts, or cross-cutting analyses addressing multiple dimensions. This categorisation facilitated subsequent thematic analysis and synthesis.
Third, key findings were extracted and analysed to identify patterns, consistencies, and contradictions across the literature. Particular attention was given to studies employing rigorous empirical methodologies, including econometric analyses of administrative data, survey-based research, and comparative international studies.
Fourth, synthesised findings were critically evaluated in relation to the research objectives, with attention to the strength of evidence, potential limitations, and implications for theory and policy. This critical evaluation forms the basis for the discussion and conclusions sections.
The literature synthesis approach offers several advantages for addressing the research questions. It enables comprehensive coverage of a multidimensional topic that spans multiple disciplines and empirical contexts. It facilitates identification of consistent patterns across diverse studies whilst also revealing important nuances and contradictions. It provides a foundation for evidence-based policy recommendations grounded in the best available research.
Limitations of this methodology include reliance on published research, which may be subject to publication bias favouring statistically significant or novel findings. Additionally, the rapidly evolving nature of AI and automation technologies means that empirical evidence may lag behind current technological capabilities and labour market developments. These limitations are acknowledged throughout the analysis.
Literature review
Theoretical foundations of technological unemployment and skill-biased change
The relationship between technological change and labour market outcomes has attracted scholarly attention for over two centuries, with perspectives evolving considerably over time. Classical economists including David Ricardo recognised that machinery could displace workers, though subsequent neoclassical theory generally emphasised compensating mechanisms through which technological progress ultimately created more jobs than it destroyed (Mokyr, Vickers and Ziebarth, 2015). The contemporary literature on AI and automation builds upon these foundational debates whilst incorporating new theoretical insights regarding the nature of human-machine complementarity.
Skill-biased technological change (SBTC) emerged as a dominant theoretical framework during the 1990s and early 2000s, positing that technological advances systematically favour workers with higher education and cognitive abilities whilst reducing demand for less-skilled labour (Acemoglu and Autor, 2011). This framework helped explain rising wage inequality and increasing returns to education observed across developed economies. However, subsequent research identified important limitations of SBTC, particularly its difficulty accounting for job polarisation—the simultaneous growth of high-skill and low-skill occupations alongside declining middle-skill employment.
The task-based approach proposed by Autor, Levy and Murnane (2003) offered a more nuanced theoretical perspective, focusing on the substitutability of specific tasks rather than aggregate skill levels. This framework distinguishes between routine tasks (both cognitive and manual) that follow explicit rules and are therefore susceptible to automation, and non-routine tasks (both abstract and manual) that require flexibility, creativity, or physical dexterity in unstructured environments. The task-based approach has proven particularly influential in contemporary analyses of AI and automation impacts.
More recent theoretical developments have emphasised the distinction between automation technologies that substitute for human labour versus augmentation technologies that complement human capabilities. Acemoglu and Restrepo (2018) formalised this distinction, demonstrating how the balance between displacement and reinstatement effects shapes aggregate employment outcomes. Their framework highlights that new technologies create new tasks requiring human labour even as they automate existing tasks, with the net employment effect depending on the relative pace of these processes.
Geographic patterns of automation exposure and impact
Contemporary research demonstrates clear spatial divergence in AI and automation impacts, with effects varying significantly across regions, cities, and local labour markets. Frank et al. (2019) provide foundational analysis showing that large, high-skill cities tend to host AI-complementary jobs and capture productivity gains from technological adoption, whilst smaller cities and rural areas disproportionately host routine, automatable work and face greater vulnerability to displacement and wage declines. This spatial polarisation reflects underlying differences in industrial composition, workforce skills, and institutional capacity.
The geographic concentration of technological impacts carries significant implications for regional inequality. National labour market statistics may mask severe localised disruptions, as aggregate employment and wage figures obscure divergent trajectories across places. Frank et al. (2019) emphasise that understanding these local impacts requires granular geographic analysis beyond regional or metropolitan aggregates.
Research on Italian regions provides additional evidence regarding the spatial dimensions of automation-induced inequality. Capello, Ciappei and Lenzi (2024) demonstrate that robotisation raises inequality through distinct mechanisms across different types of places. In non-metropolitan manufacturing areas, automation primarily operates through job loss for low-skilled workers, whilst in cities it drives inequality through growth of high-skill jobs and widening between-group wage gaps. This dual mechanism highlights how the same technology can produce inequality through different channels depending on local economic structures.
The spatial concentration of automation vulnerability has historical antecedents in earlier waves of industrial restructuring. Regions that experienced deindustrialisation during the late twentieth century often exhibit persistent economic disadvantage, and many of these same places now face heightened exposure to automation of remaining manufacturing and routine service employment (Autor, Dorn and Hanson, 2016). This path dependency suggests that automation impacts may compound rather than counteract existing spatial inequalities.
Urban-rural differences represent a particularly salient dimension of spatial variation. Urban areas, particularly large metropolitan regions, tend to specialise in knowledge-intensive activities that complement AI capabilities, whilst rural areas and smaller towns more frequently host routine activities susceptible to automation. However, this generalisation masks considerable variation, with some smaller cities and rural areas developing niches in high-technology sectors whilst some urban neighbourhoods experience concentration of vulnerable low-skill service employment.
Sectoral and occupational patterns of technological exposure
The literature reveals distinct patterns of AI and automation exposure across industrial sectors and occupational categories. A fundamental distinction emerges between physical automation through robotics, which primarily affects manufacturing and routine manual tasks, and cognitive automation through AI, which increasingly affects knowledge work and professional services.
Routine, manual, and middle-skill jobs in manufacturing, retail, and basic services consistently emerge as most exposed to automation and job displacement. Sultana et al. (2024) document the macroeconomic implications of automation for these sectors, finding significant negative employment effects concentrated among workers performing repetitive, codifiable tasks. Shen and Zhang (2024) similarly identify manufacturing as facing substantial displacement, whilst also noting the role of virtual agglomeration in mediating AI’s employment effects. Yolusever (2025) and Du (2024) provide additional evidence regarding the concentration of unemployment risk in routine occupations.
Cognitive, high-skill work in technology, finance, and advanced services exhibits different patterns of AI exposure, characterised more by augmentation and new task creation than displacement. Tyson and Zysman (2022) analyse how AI transforms work in these sectors, finding that whilst some cognitive tasks become automated, new human roles emerge around AI system development, oversight, and application. Sultana et al. (2024) note that workers in these sectors more frequently benefit from AI-induced productivity gains, though often with increased work intensity and changing skill requirements.
Acemoglu et al. (2022) provide detailed evidence on AI-related job vacancies, demonstrating how artificial intelligence creates demand for new occupational categories whilst also revealing the uneven distribution of these opportunities across firms and regions. Their analysis of online vacancy data shows that AI-exposed establishments tend to be larger, more productive, and concentrated in specific metropolitan areas.
Lankisch, Prettner and Prskawetz (2019) contribute theoretical and empirical analysis of how robots affect wage inequality across sectors, finding that automation tends to reduce labour demand in routine occupations whilst increasing demand for high-skill workers who program, maintain, and supervise automated systems. This pattern produces both job displacement and wage compression at the lower end of the occupational distribution.
Research by Webb (2019) provides one of the most comprehensive analyses of AI versus automation exposure, developing occupation-level measures based on the overlap between job tasks and patented technologies. This research demonstrates that AI exposure differs substantially from earlier forms of automation exposure, affecting a broader range of occupations including some high-skill professional roles previously considered insulated from technological substitution.
Jaccoud (2025) offers particularly nuanced analysis of how robots and AI differentially affect wage inequality within occupations. This research finds that robot exposure tends to compress wages at the bottom of occupational wage distributions, whilst AI exposure increases within-occupation wage dispersion, particularly at the top. This divergent pattern suggests that AI may operate more through skill-biased mechanisms within occupations rather than between-occupation displacement effects.
Marguerit (2025) examines whether AI development augments or automates labour, finding evidence for both mechanisms operating simultaneously but with augmentation effects more prominent for workers with skills complementary to AI capabilities. Grant and Üngör (2024) similarly document how AI interacts with twenty-first century skills to shape wage inequality, finding that workers combining formal education with AI-relevant competencies command substantial wage premiums.
Educational attainment and the automation-skill gradient
Educational attainment emerges as perhaps the most consequential individual-level factor mediating AI and automation impacts. The literature consistently demonstrates that automation increases returns to higher education and widens wage gaps between high-skilled and low-skilled workers, extending patterns observed with earlier forms of technological change.
Prettner and Strulik (2020) provide comprehensive analysis of how innovation and automation affect inequality, documenting the mechanisms through which technological change raises returns to education. Their research demonstrates that workers with higher educational attainment are better positioned to perform the non-routine cognitive tasks that remain human domains even as routine work becomes automated. Additionally, education facilitates occupational mobility, enabling workers to transition away from automatable jobs toward emerging roles.
Lankisch, Prettner and Prskawetz (2019) extend this analysis to examine explicitly how robots affect wage inequality through the education channel. Their findings confirm that automation systematically disadvantages workers with lower educational attainment, who disproportionately occupy routine roles susceptible to technological substitution. Conversely, workers with tertiary education increasingly specialise in tasks complementary to automated systems, capturing wage premiums from technology-enhanced productivity.
Beyond general educational attainment, specific skills related to AI and digital technologies confer additional labour market advantages. Grant and Üngör (2024) document an “AI skill premium” whereby workers with competencies in artificial intelligence, data science, and related fields earn substantially more than peers with comparable traditional educational credentials. This premium reflects the scarcity of AI-relevant skills relative to employer demand, and it may either expand or contract as AI training becomes more widely available.
Yolusever (2025) examines how education shapes vulnerability to AI-induced displacement, finding that workers with lower educational attainment face substantially higher unemployment risk as AI capabilities expand. However, this research also identifies potential for education policy to mitigate these effects, suggesting that investments in adult learning and reskilling can reduce displacement risk for workers willing and able to acquire new competencies.
The relationship between education and automation vulnerability intersects with other demographic and social characteristics. Sultana et al. (2024) note that disability status may compound educational disadvantage, as disabled workers often face both lower educational attainment on average and additional barriers to occupational mobility. Omri, Omri and Afi (2025) provide detailed analysis of AI impacts on unemployment for people with disabilities, finding that educational attainment and governance quality significantly moderate these effects. Their research suggests that inclusive education policies combined with effective regulatory frameworks can substantially reduce automation-related employment penalties for disabled workers.
Mechanisms linking automation to inequality
The literature identifies several distinct mechanisms through which AI and automation generate labour market inequality across geographic, sectoral, and educational dimensions. Understanding these mechanisms is essential for designing effective policy responses.
First, direct displacement occurs when automated systems substitute for human workers performing specific tasks, reducing labour demand in affected occupations. This mechanism operates most directly in routine manufacturing and service roles, where codifiable tasks can be performed by machines at lower cost. Displaced workers may face unemployment or transition to lower-paying alternative employment, generating adverse wage and employment effects concentrated among vulnerable populations.
Second, wage compression at the lower end of the labour market results from increased competition among workers displaced from automatable jobs. As workers shift toward remaining low-skill service occupations that resist automation, labour supply in these roles increases relative to demand, depressing wages. This mechanism contributes to widening gaps between low and middle wages even as it may compress wages within the low-skill segment.
Third, skill premiums for AI-complementary work arise from increased demand for workers capable of developing, implementing, and leveraging automated systems. Workers with relevant technical skills command higher wages as employers compete for scarce talent. This mechanism operates both between occupations (raising wages in high-skill roles relative to others) and within occupations (raising wages for top performers within given roles).
Fourth, productivity gains and rent capture occur when automation enables substantial efficiency improvements, generating economic surplus. The distribution of these gains among capital owners, high-skill workers, and consumers shapes inequality outcomes. Evidence suggests that capital owners and workers with AI-complementary skills capture disproportionate shares of automation-generated productivity gains, whilst displaced workers and their communities bear concentrated costs.
Fifth, agglomeration effects concentrate both the benefits and costs of automation geographically. High-skill workers and innovative firms cluster in specific places, creating knowledge spillovers and thick labour markets that reinforce local advantages. Conversely, places that lose employment to automation may experience cumulative disadvantage as skilled workers depart and local demand declines.
Discussion
Integrating evidence across dimensions of inequality
The synthesised evidence reveals AI and automation as powerful forces amplifying existing labour market inequalities across geographic, sectoral, and educational dimensions. Rather than operating independently, these dimensions interact in ways that concentrate disadvantage among specific populations and places whilst conferring compounded advantages on others. This pattern of cumulative causation poses significant challenges for policy efforts to promote inclusive technological transitions.
Geographic inequality emerges not merely as a backdrop but as an active mechanism through which automation effects are channelled. The concentration of AI-complementary employment in large, prosperous cities creates self-reinforcing dynamics whereby skilled workers migrate toward opportunity whilst declining regions lose both jobs and human capital. This spatial sorting amplifies the consequences of technological change beyond direct displacement effects, potentially contributing to long-term regional divergence that undermines both economic efficiency and social cohesion.
The evidence regarding Frank et al.’s (2019) observation that national statistics mask localised disruption carries important implications for labour market monitoring and policy design. Aggregate employment figures may suggest manageable adjustment to technological change even as specific communities experience severe dislocation. Effective policy response requires granular geographic analysis capable of identifying emerging problems before they become entrenched.
Sectoral patterns of automation exposure reflect the fundamental distinction between routine and non-routine work identified in the task-based theoretical framework. The evidence strongly supports this distinction whilst adding important nuance regarding the different inequality effects of robots versus AI. Jaccoud’s (2025) finding that robots compress wages within occupations whilst AI increases wage dispersion suggests that these technologies operate through distinct mechanisms requiring differentiated policy responses.
The sectoral analysis also highlights the evolving frontier of automation capability. Whilst earlier automation primarily affected manufacturing, current and emerging AI systems increasingly encroach upon knowledge work and professional services. Webb’s (2019) analysis of AI exposure across occupations suggests that the population of workers vulnerable to technological displacement may expand substantially beyond those affected by earlier automation waves. This prospect heightens the importance of educational and training systems capable of equipping workers for ongoing occupational transitions.
Educational attainment’s role as a key mediator of automation impacts underscores both challenges and opportunities for policy intervention. On one hand, the strong relationship between education and automation resilience raises concerns about workers with limited educational opportunities, who face compounding disadvantages in increasingly skill-biased labour markets. On the other hand, the malleability of skills provides a potential lever for policy intervention, as investments in education and training could reduce automation vulnerability among currently exposed populations.
Critical evaluation of the evidence base
Whilst the synthesised evidence provides consistent support for the overall pattern of unequal automation impacts, several limitations warrant acknowledgement. First, much of the empirical research examines regions or sectors where automation has been relatively advanced, potentially limiting generalisability to contexts at earlier stages of technological adoption. The relationship between automation exposure and labour market outcomes may differ as technologies mature and diffuse more widely.
Second, the distinction between AI and robot exposure, whilst analytically useful, becomes increasingly difficult to maintain as technologies converge. Contemporary automation systems often combine physical robotics with AI-based perception and decision-making, suggesting that separate analyses of robot and AI impacts may miss important interaction effects.
Third, most empirical studies focus on developed economies, particularly the United States and Western Europe. The transferability of findings to developing country contexts, where labour market institutions and sectoral compositions differ substantially, remains uncertain. Global supply chains mean that automation in one location may affect employment elsewhere, requiring international perspectives that much of the current literature lacks.
Fourth, the time horizon of analysis affects interpretation of automation impacts. Studies examining short-term effects may capture displacement whilst missing subsequent job creation; conversely, long-run analyses may underestimate adjustment costs borne by affected workers and communities during transitional periods. The appropriate temporal frame for assessing automation impacts remains contested.
Achievement of research objectives
The analysis provides substantial evidence addressing each of the specified research objectives. Regarding geographic variations (Objective 1), the literature synthesis demonstrates clear spatial divergence, with large cities and prosperous regions capturing automation benefits whilst smaller cities and rural areas bear disproportionate displacement risks. The evidence supports characterisation of AI and automation as spatially concentrating rather than dispersing economic activity, with significant implications for regional inequality.
Regarding sectoral patterns (Objective 2), the analysis documents distinct trajectories for routine versus non-routine occupations and for sectors differently positioned relative to AI capabilities. Manufacturing and basic services face displacement whilst knowledge-intensive sectors experience augmentation, though the expanding frontier of AI capability may extend automation exposure to previously protected occupations.
Regarding educational mediation (Objective 3), the evidence strongly supports the role of educational attainment in shaping automation vulnerability and benefit capture. Workers with higher education and AI-relevant skills command significant premiums, whilst those with limited education face elevated displacement and wage penalty risks.
Regarding inequality implications (Objective 4), the synthesis documents automation’s contribution to multiple dimensions of labour market inequality: between-group wage gaps, within-occupation wage dispersion, and spatial economic divergence. These dimensions interact cumulatively, suggesting that technological change may compound existing disadvantage.
Regarding policy implications (Objective 5), the evidence points toward the importance of education and training investments, place-based support for affected communities, and governance frameworks capable of shaping automation trajectories. However, the literature provides less specific guidance regarding the design and implementation of such interventions, suggesting need for continued research and policy experimentation.
Implications for policy and practice
The documented patterns of unequal automation impact carry significant implications for policy across multiple domains. Education and training systems face the challenge of equipping both new workforce entrants and existing workers with skills complementary to AI capabilities. This requires not only expanded access to higher education but also transformation of curricula to emphasise the non-routine, creative, and interpersonal competencies that resist automation.
Place-based policies may be essential for addressing the geographic concentration of automation impacts. Policies supporting diversification of local economies, investments in infrastructure and amenities that attract skilled workers, and transitional support for displaced workers could mitigate the cumulative disadvantage facing automation-vulnerable regions. However, such policies require sustained commitment and significant resources, and their effectiveness in countering powerful agglomeration dynamics remains uncertain.
Labour market institutions including minimum wages, social insurance systems, and worker voice mechanisms may require adaptation to address automation-related challenges. Strong minimum wages could prevent excessive compression of wages at the lower end of the labour market, whilst expanded social insurance could provide security during occupational transitions. Worker representation in technology adoption decisions could help ensure that productivity gains are broadly shared.
Governance of AI development and deployment represents an emerging policy frontier. Omri, Omri and Afi (2025) find that governance quality significantly moderates automation impacts on vulnerable populations, suggesting that regulatory frameworks shaping how technologies are developed and implemented can influence distributional outcomes. Policies encouraging AI applications that augment rather than simply substitute for human capabilities could help steer technological trajectories toward more inclusive outcomes.
Conclusions
This dissertation has examined the differential impacts of artificial intelligence and automation on labour markets across geographic areas, industrial sectors, and educational attainment levels. Through comprehensive synthesis of contemporary peer-reviewed research, the analysis provides substantial evidence that AI and automation amplify existing inequalities, rewarding AI-complementary skills and high-productivity regions whilst concentrating displacement among routine, low-skill jobs and already vulnerable local labour markets.
The research objectives have been achieved through systematic analysis of evidence regarding each dimension of inequality. Geographic analysis demonstrates that large, high-skill cities capture productivity benefits whilst smaller cities and rural areas bear disproportionate displacement risks. Sectoral analysis reveals distinct patterns for routine versus non-routine occupations, with robots and AI operating through different inequality mechanisms. Educational analysis confirms that attainment strongly mediates automation vulnerability, with significant premiums for workers possessing AI-complementary skills.
The significance of these findings extends beyond academic understanding to practical implications for policy and practice. The documented patterns of cumulative disadvantage—whereby technological change compounds existing inequalities across space, sector, and education—underscore the urgency of policy responses promoting more inclusive technological transitions. Education and training investments, place-based support programmes, adaptive labour market institutions, and governance frameworks shaping technology development all represent potential intervention points.
Future research should address several gaps identified in the current literature. Longer-term studies tracking the dynamic adjustment of labour markets to technological change would help distinguish transitory displacement from permanent restructuring. Comparative international analysis would illuminate how institutional differences across countries shape automation impacts. Research examining the interaction between AI and automation exposure and other dimensions of disadvantage, including disability, ethnicity, and gender, would support more targeted policy responses. Finally, evaluation of specific policy interventions would build the evidence base regarding effective responses to automation-related labour market challenges.
The challenge of ensuring that AI and automation benefit workers and communities broadly rather than concentrating gains among the already advantaged represents one of the defining policy questions of our era. This dissertation contributes to understanding the mechanisms and patterns of unequal technological impact, providing a foundation for the continued research and policy innovation needed to achieve more inclusive outcomes.
References
Acemoglu, D. and Autor, D. (2011) ‘Skills, tasks and technologies: Implications for employment and earnings’, in Ashenfelter, O. and Card, D. (eds.) *Handbook of Labor Economics*, Volume 4B. Amsterdam: Elsevier, pp. 1043–1171.
Acemoglu, D. and Restrepo, P. (2018) ‘Artificial intelligence, automation and work’, in Agrawal, A., Gans, J. and Goldfarb, A. (eds.) *The Economics of Artificial Intelligence: An Agenda*. Chicago: University of Chicago Press, pp. 197–236.
Acemoglu, D., Autor, D., Hazell, J. and Restrepo, P. (2022) ‘Artificial intelligence and jobs: Evidence from online vacancies’, *Journal of Labor Economics*, 40(S1), pp. S293–S340. https://doi.org/10.1086/718327
Autor, D.H. (2015) ‘Why are there still so many jobs? The history and future of workplace automation’, *Journal of Economic Perspectives*, 29(3), pp. 3–30.
Autor, D.H., Dorn, D. and Hanson, G.H. (2016) ‘The China shock: Learning from labor-market adjustment to large changes in trade’, *Annual Review of Economics*, 8(1), pp. 205–240.
Autor, D.H., Levy, F. and Murnane, R.J. (2003) ‘The skill content of recent technological change: An empirical exploration’, *Quarterly Journal of Economics*, 118(4), pp. 1279–1333.
Brynjolfsson, E. and McAfee, A. (2014) *The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies*. New York: W.W. Norton.
Capello, R., Ciappei, S. and Lenzi, C. (2024) ‘Unveiling the automation—wage inequality nexus within and across regions’, *The Annals of Regional Science*, 73(4), pp. 1729–1756. https://doi.org/10.1007/s00168-024-01317-7
Du, J. (2024) ‘The impact of artificial intelligence adoption on employee unemployment: A multifaceted relationship’, *International Journal of Social Sciences and Public Administration*, 2(3), pp. 304–311. https://doi.org/10.62051/ijsspa.v2n3.45
Frank, M., Autor, D., Bessen, J., Brynjolfsson, E., Cebrián, M., Deming, D., Feldman, M., Groh, M., Lobo, J., Moro, E., Wang, D., Youn, H. and Rahwan, I. (2019) ‘Toward understanding the impact of artificial intelligence on labor’, *Proceedings of the National Academy of Sciences*, 116(14), pp. 6531–6539. https://doi.org/10.1073/pnas.1900949116
Grant, R. and Üngör, M. (2024) ‘The AI revolution with 21st century skills: Implications for the wage inequality and technical change’, *Scottish Journal of Political Economy*, 71(5), pp. 640–668. https://doi.org/10.1111/sjpe.12395
Jaccoud, F. (2025) ‘Robots & AI exposure and wage inequality: A within occupation approach’, *Eurasian Business Review*, 15(1), pp. 1–28. https://doi.org/10.53330/eajl3597
Lankisch, C., Prettner, K. and Prskawetz, A. (2019) ‘How can robots affect wage inequality?’, *Economic Modelling*, 81, pp. 161–169. https://doi.org/10.1016/j.econmod.2018.12.015
Marguerit, D. (2025) ‘Augmenting or automating labor? The effect of AI development on new work, employment, and wages’, *SSRN Electronic Journal*. https://doi.org/10.2139/ssrn.5169611
Mokyr, J., Vickers, C. and Ziebarth, N.L. (2015) ‘The history of technological anxiety and the future of economic growth: Is this time different?’, *Journal of Economic Perspectives*, 29(3), pp. 31–50.
Omri, A., Omri, H. and Afi, H. (2025) ‘Exploring the impact of AI on unemployment for people with disabilities: Do educational attainment and governance matter?’, *Frontiers in Public Health*, 13, 1559101. https://doi.org/10.3389/fpubh.2025.1559101
Prettner, K. and Strulik, H. (2020) ‘Innovation, automation, and inequality: Policy challenges in the race against the machine’, *Journal of Monetary Economics*, 116, pp. 249–265. https://doi.org/10.1016/j.jmoneco.2019.10.012
Shen, Y. and Zhang, X. (2024) ‘The impact of artificial intelligence on employment: The role of virtual agglomeration’, *Humanities and Social Sciences Communications*, 11, 188. https://doi.org/10.1057/s41599-024-02647-9
Snyder, H. (2019) ‘Literature review as a research methodology: An overview and guidelines’, *Journal of Business Research*, 104, pp. 333–339.
Sultana, F., Talpur, U., Iqbal, M., Ali, A. and Memon, K. (2024) ‘The macroeconomic implications of automation and AI on labor markets and employment’, *The Critical Review of Social Sciences Studies*, 2(2), pp. 89–103. https://doi.org/10.59075/f9hdkk61
Tyson, L. and Zysman, J. (2022) ‘Automation, AI & work’, *Daedalus*, 151(2), pp. 256–271. https://doi.org/10.1162/daed_a_01914
Webb, M. (2019) ‘The impact of artificial intelligence on the labor market’, *Stanford University Working Paper*. https://doi.org/10.2139/ssrn.3482150
Yolusever, A. (2025) ‘AI and automation: Reshaping the labor market’, *Biga İktisadi ve İdari Bilimler Fakültesi Dergisi*, 6(1), pp. 45–62. https://doi.org/10.70754/biibfd.1594580
