Figure 1
The illustration is a three-column table comparing “Human intelligence”, “Augmented intelligence”, and “Artificial intelligence” for the task of choosing J E L (Journal of Economic Literature) codes. Each column includes a set of bulleted points describing unique features: Human intelligence: Choosing J E L codes manually with limited cognitive abilities and under bounded rationality. Learning of the J E L codes guide is a costly and time-consuming task. Subjective beliefs and expectations may matter in the choice of J E L codes. Can flexibly adapt to changes in the J E L classification codes system (e.g., introduction of new J E L codes). Augmented intelligence: A I systems can complement and augment researchers’ abilities in choosing appropriate J E L codes. Possibility to semi-automate better-informed choices of J E L codes with the synergy of human intelligence and A I systems. Artificial intelligence: Choosing J E L codes objectively based on patterns of training data and machine learning. Possibility to automate the choice of J E L codes (what should an A I system optimize in the choice of J E L codes)? Generative A I systems (chatbots) can provide reasoning why specific J E L codes are appropriate. May not be as flexible as human intelligence in adapting to changes in the J E L classification codes system (e.g., training data related to novel J E L codes accumulates over time).Comparing human intelligence, augmented intelligence and AI in the choice of JEL codes