From Job Displacement to Task Reallocation: Evidence from Temporal Analysis of Data Science Job Postings

Abstract Recent advances in artificial intelligence (AI), particularly generative AI, have increased public concern regarding the potential displacement of data science professionals. Popular discourse frequently frames AI as a direct substitute for analytical labor, fueling fears of widespread job loss. However, jobs are bundles of tasks, and technological change often reshapes work internally rather than removing occupations outright. This study adopts a task oriented perspective to examine whether AI diffusion corresponds with job displacement or, instead, task reallocation within the data science profession. Using a comparative analysis of U.S. based Data Scientist job postings from a pre-diffusion baseline derived from a Glassdoor scraped dataset (published in the late 2010s) and a post-diffusion dataset of 2023 postings curated by Luke Barousse, this paper analyzes changes in task and skill related language as proxies for evolving employer expectations. The analysis indicates substantial shifts in the composition of emphasized skills and responsibilities rather than a collapse in demand. In particular, several traditional tools and routine signals have declined in relative share, while signals related to orchestration and modern production workflows have increased. These findings are consistent with a task reallocation framework and motivate task level measurement as a complement to occupation level narratives. Keywords artificial intelligence, labor markets, task reallocation, job postings, data science