Reproducing and Extending Experiments in Behavioral Strategy with Large Language Models

Two prominent approaches have emerged to advance our understanding of these microfoundations of strategy: computational work and human lab experiments. Agent-based computational simulations have sharpened our understanding of performance and learning consequences stemming from differences in individuals' cognition (Csaszar and Levinthal 2016, Gavetti and Levinthal 2000, Knudsen and Srikanth 2014, Winter et al. 2007). Additionally, scholars have increasingly designed experiments to study human responses within various tasks, such as searching for high-performing alternatives in unknown decision-spaces (Bergenholtz et al. 2023, Billinger et al. 2014, 2021, Richter et al. 2023), self-selecting into specific organizational tasks (Raveendran et al. 2022), exhibiting organizational voting behavior (Piezunka and Schilke 2023), and making innovation choices in response to different organizational contingencies (Klingebiel 2022). Despite significant strides, a key challenge in advancing behavioral strategy lies in building and testing theories of individual-level cognition and its effects on the revealed decisions that our field typically focuses on. More theoretical development and empirical testing are needed to understand when and why decision-makers follow particular heuristics in specific situations, and what task factors influence their cognitive processes.