The paper addresses the ongoing debate surrounding the impact of AI on employment, particularly in the context of China's rapidly developing economy and large workforce. Sustainable Development Goal 8 (full and productive employment) provides the backdrop for this research. The introduction notes that while technological advancements often lead to job displacement, they also create new opportunities. The study aims to empirically assess the net effect of AI, represented by industrial robot adoption, on employment in China. The authors acknowledge the conflicting views on the topic – some arguing that AI will lead to widespread unemployment, others that it will create new jobs. China's unique position as a developing nation with substantial industrial growth and labor market dynamics makes it a particularly relevant case study. The research will explore not only the overall impact but also the heterogeneous effects across different demographics and industries. The role of virtual agglomeration, a phenomenon driven by the digital economy, will also be examined as a potential mechanism driving employment growth.

The literature review examines existing arguments on AI's impact on employment. Three main perspectives are identified: (1) AI's job creation effect, emphasizing improved productivity, increased consumption, and the creation of new, complex tasks; (2) AI's destructive and substitution effects, highlighting the potential for job displacement, especially for low-skilled workers; and (3) the nuanced perspective that considers both the creation and destruction of jobs, acknowledging the complexities of technological progress. The review discusses various economic theories, including Schumpeter's innovation theory and Marx's analysis of technological unemployment. The authors highlight the inconsistency in existing research findings regarding the net effect of AI on employment and the need for further research, particularly in developing countries like China.

The study uses panel data from 30 Chinese provinces (excluding Hong Kong, Macau, Taiwan, and Xizang) from 2006 to 2020. The core explanatory variable is the installation density of industrial robots, serving as a proxy for AI adoption. The explained variable is the employment scale in the manufacturing sector. The authors address the challenge of data compatibility between IFR's industry classification and China's, using a share method to merge and aggregate data. A two-way fixed-effect model is employed to control for unobserved heterogeneity across provinces and over time. The Hausman and F tests are used to justify the choice of fixed effects. To address potential endogeneity concerns (a bidirectional relationship between labor demand and robot adoption), the study uses a two-stage least squares (2SLS) method, employing industrial electricity prices as an instrumental variable. This choice is based on the argument that electricity prices influence robot adoption but do not directly affect the labor market. Mediating variables (labor productivity, capital deepening, division of labor refinement, and virtual agglomeration) are included to analyze the mechanisms through which AI affects employment. Robustness checks are conducted by replacing explained variables, adding control variables, and using a machine learning approach to assess the importance of AI in employment growth. The study also investigates the heterogeneity of AI's impact across gender and industry attributes.

The study's key findings are robust across different model specifications and robustness checks. The two-way fixed-effect model and the 2SLS estimations consistently show a positive and significant relationship between AI adoption (as measured by industrial robot density) and employment scale. This implies that the job creation effect of AI outweighs the job displacement effect in the Chinese context. The magnitude of the positive effect varies across different model specifications. The analysis of mediating variables reveals that AI promotes employment by improving labor productivity, deepening capital, and refining the division of labor. The study also highlights the significant role of virtual agglomeration, which facilitates knowledge spillover and creates new employment opportunities. The positive effects of AI on employment are not uniform across all groups; women and workers in labor-intensive industries seem to benefit disproportionately. Results from the robustness checks, including the replacement of variables, inclusion of additional control variables, and machine learning techniques, further support the overall positive effect of AI on employment.

The findings challenge the common perception of AI as a primary source of job displacement. The study demonstrates that in the Chinese context, AI's job creation effects currently dominate the substitution effects. This could be attributed to several factors, including China's relatively low labor costs, the still-nascent stage of AI adoption in many sectors, and the emergence of new job opportunities in related fields. The heterogeneous impact across gender and industry highlights the importance of considering the broader context of labor market dynamics when assessing AI's effects. The significant role of virtual agglomeration underscores the importance of digital infrastructure and interconnectedness in shaping employment outcomes in the age of AI. The results suggest a need to rethink the narrative around AI and employment, shifting away from simplistic views of job displacement towards a more nuanced understanding of the complex interplay between technology, economic growth, and labor market dynamics.

The study concludes that AI, as measured by industrial robot adoption, has a positive net effect on employment in China. This positive effect operates through various mechanisms, including increased labor productivity, capital deepening, division of labor refinement, and the expansion of virtual agglomeration. The study's findings provide valuable insights into the impact of AI on employment in a developing country context and highlight the importance of proactive policy interventions to maximize the benefits of AI while mitigating potential risks. Future research should explore AI's micro-level impacts, conduct policy evaluations, and investigate spatial spillover effects.

The study acknowledges several limitations. The analysis is based on macro-level data, limiting insights into micro-level dynamics. While the methods employed address endogeneity concerns to some degree, true causal inference is not fully established. The study's conclusions are specific to the Chinese context and the sample period; generalizability to other countries or time periods requires further investigation. Future research should include more granular data, policy experiments, and cross-country comparisons to address these limitations.