Will robots or artificial intelligence take our jobs? At the center of the debate about the future of work is “technological unemployment”, a term that has a seemingly simple definition but has in fact been used and defined differently by economists. In this paper, we explore how economists have discussed the potential for new techniques to replace workers since Aristotle, and how they have defined and conceived of technological unemployment over the past century. We begin with a detailed analysis of classic texts on this topic, from ancient times to the 20th century. To capture changes in the research frontier, we quantitatively and qualitatively analyze all 153 articles that mention the term “technological unemployment” in twelve major economics journals, including the top five, since their inception. We then use the 19 editions of Paul Samuelson’s seminal textbook and a cross-section of 43 economics textbooks from the 2000s and 2010s to observe the state of discourse and changes in economics pedagogy. Our analysis shows that economists have used a range of definitions in their discussions of technological unemployment, and most definitions are brief and imprecise. Economics textbooks notably omit technological unemployment in their discussions of the relationship between technological change and employment, despite the continuing interest in the topic in the academic literature. Nonetheless, we find a surprising consensus in our corpus that technological change may cause unemployment. Over time, the debate around technological unemployment has become narrower and more technical, but also more heated during historical periods of technological anxiety. We suggest that the adoption of a clear definition with specific temporal and scale modifiers could clarify theoretical debates and improve the precision of future empirical research on the topic, which will allow economists to speak directly to public and policy concerns
