Does AI Cheapen Talk? Theory and Evidence From Global Entrepreneurship and Hiring

Abstract

Screening human capital based on signals such as job applications or entrepreneurial pitches is crucial for organizations. Signals are informative insofar as they are costly. Generative AI (GAI) complicates screening by lowering the cost of producing impressive signals. We model the informational effects of GAI, showing that applicants' use of GAI can increase-but also decrease-an evaluator's screening mistakes. This result depends on how GAI affects experts' signals compared to non-experts'. Using experiments in hiring and startup investing, we estimate that the adoption of GAI (ChatGPT) lowers employers' and startup investors' screening accuracy by 4-9%. Consistent with our model, we also find that the use of GAI in some settings improves screening accuracy (in our case, non-English-speaking countries). These results show that GAI can profoundly shape screening accuracy.

Keywords: Screening, Artificial Intelligence, Entrepreneurship, Human Capital

JEL Classification: M10, M13, O33