Does AI Cheapen Talk? Evidence From Global Startup and Hiring Contexts

35 Pages Posted: 22 Jan 2024 Last revised: 6 Feb 2024

See all articles by Nataliya Langburd Wright

Nataliya Langburd Wright

Columbia University - Columbia Business School, Management

Bo Cowgill

Columbia University - Columbia Business School

Date Written: January 21, 2024

Abstract

New AI technology now makes it easier to send signals of quality, whether it be in a job application or a startup pitch. This study is the first to assess how AI changes the informational value of such signals. We theorize that AI could either increase or decrease information transmission, depending on whether AI complements or substitutes prior signaling ability. In experiments in entrepreneurship and hiring settings, we find that generative AI (ChatGPT) lowers information transmission by 2% on average and increases demand for costlier, more objective signals both by senders and receivers. Information loss is stronger for senders with lower ex-ante signaling ability for our tasks, including senders from lower-income countries. These results are consistent with a model of AI as a substitute for ex-ante signaling ability.

Keywords: Artificial Intelligence, Digital Economy, Entrepreneurial Strategy, Entrepreneurship, Human Capital

JEL Classification: M10, M13, O33

Suggested Citation

Wright, Nataliya and Cowgill, Bo, Does AI Cheapen Talk? Evidence From Global Startup and Hiring Contexts (January 21, 2024). Columbia Business School Research Paper Forthcoming, Available at SSRN: https://ssrn.com/abstract=4702114 or http://dx.doi.org/10.2139/ssrn.4702114

Nataliya Wright (Contact Author)

Columbia University - Columbia Business School, Management ( email )

3022 Broadway
New York, NY 10027
United States

Bo Cowgill

Columbia University - Columbia Business School ( email )

3022 Broadway
New York, NY 10027
United States

Do you have negative results from your research you’d like to share?

Paper statistics

Downloads
146
Abstract Views
461
Rank
358,119
PlumX Metrics