Artificial Intelligence and the Labor Market: Measuring Automation Potential, Human Advantage, and the Limits of Current Metrics

Abstract

As artificial intelligence continues to reshape the labor market, the ability to accurately measure and forecast its impact on the workforce has become a crucial focus. Policymakers, economists, and artificial intelligence researchers have turned to a wide range of approaches to inform future policy decisions—from asking humans about their views on job automatability to prompting large language models (LLMs) to predict automatability. This thesis argues that neither human-based nor LLM-based approaches to automation forecasting are currently reliable enough to inform policy, and that the most pressing need is standardized, enforceable data collection. An in-depth analysis of the literature, and two demonstrative analyses, were conducted to examine the effectiveness and limitations of existing methods, suggesting gaps in both current forecasting and data analysis approaches. Utilizing these results as well as enlisting the support of existing federal, state, and international policy responses, this thesis concludes that meaningful labor market protection requires a more transparent and detailed data collection. Without it, neither economists nor policymakers can accurately identify the ways in which AI is already reshaping the labor market.