A History of Rationality in Economics: a Computational Approach
Working paper
Abstract
This paper offers a concrete, method-driven account of how quantitative techniques can illuminate the history of a capacious concept in economics—rationality. We assemble a large full-text corpus (nearly 300,000 economics articles since 1900) and pair it with structured citation data. On the textual side, we use large language models (LLMs) to trace semantic shifts over more than a century and to identify a corpus of texts centered on rationality, which we then analyze with bibliometrics and network methods. Our objectives are twofold: first, to provide a concrete methodological account that foregrounds the practical choices—from data collection to interpretation—that shape corpus construction and the reading of results; second, to show that unsupervised models, when combined with robustness checks and close reading, can serve as powerful tools for both confirmation and discovery in the history of economics. To ensure transparency and reuse, we release a Shiny application that exposes the indicators we rely on for qualitative interpretation and makes the workflow auditable and replicable.