Machine learning and the labor market: a portrait of occupational and worker inequities in Canada

INTRODUCTION: Machine learning (ML), an artificial intelligence (AI) subfield, is increasingly used by Canadian workplaces. Concerningly, the impact of ML may be inequitable and contribute to social and health inequities in the working population. The aim of this study is to estimate the number of workers in occupations with high, medium, and low ML exposure and describe differences in exposure according to occupational and worker sociodemographic factors. METHODS: Canadian occupations were scored according to the extent to which they were made up of job tasks that could be performed by ML. Eight years of data from Canada's Labour Force Survey were pooled and the number of Canadians in occupations with high, medium, or low exposure to ML were estimated. The relationship between hourly wages, educational attainment, and job skill, training and experience requirements, and ML exposure was examined using multinomial models that were stratified by gender. RESULTS: Approximately 5.7 million Canadians are working in occupations characterized by high ML exposure. Women workers and workers with a college or bachelor's degree and in occupations with lower job skills requirements made up a greater proportion of workers in occupations with high ML exposure. Multinomial models indicated gender differences in the relationship between independent variables and ML exposure. Among men, higher educational attainment and hourly wages were associated with high occupational ML exposure. However, among women, higher educational attainment and hourly wages were associated with low occupational ML exposure. CONCLUSION: ML exposure is segmented according to occupational and worker sociodemographic characteristics and has the potential to widen inequities in the working population. ML may have a gendered effect and disproportionately impact certain groups of women when compared to men. We provide a critical evidence base to inform strategic responses that ensure inclusion in a working world where ML is commonplace