The Effects of AIGC Adoption on Competition and Gender Inequality in the Labor Market: A Behavioral Experiment Based on ChatGPT

How to cite

QIN, Lisha; ZHANG, Xu. 2024. "The Effects of AIGC Adoption on Competition and Gender Inequality in the Labor Market: A Behavioral Experiment Based on ChatGPT." AEA RCT Registry. AEARCTR-0014578. https://doi.org/10.1257/rct.14578-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Lisha QIN* , Xu ZHANG

* Primary investigator

Registry ID
AEARCTR-0014578
DOI
https://doi.org/10.1257/rct.14578-1.0
First registered
Last updated
Status
In development
Trial dates
2024-11-01 – 2026-01-01
Countries
Private
JEL codes
C91, D91
Registry record
http://www.socialscienceregistry.org/trials/14578

Abstract

The rapid spread of Artificial Intelligence-Generated Content (AIGC) tools, such as ChatGPT, is expected to significantly enhance labor productivity. However, their impact on different populations in the labor market remains unclear. For instance, previous research has shown that there is a significant difference in competitive preferences between men and women, which is closely related to the gender wage gap in the labor market. AIGC tools, serve as a disruptive technology, possibly serve as an additional resource that change the existing gender wage gap. On the one hand, AIGC tools may reduce gender differences in productivity in certain industries with simple tasks, but on the other hand, they may exacerbate gender inequality due to inherent differences in confidence and technology acceptance between the genders. Thus, the impact of AIGC tools on labor market competition and gender inequality is still uncertain. Taking ChatGPT as an example, this study conducts a behavioural experiment to investigate whether the availability and suitability of ChatGPT significantly affects competitive preferences and task performance, and whether these effects differ by gender. In detail, participants will experience the stages following the classic competition willingness experiment by Niederle & Vesterlund (2007), and their competition willingness will be measured by the choice-based scheme on whether to compete under different payment schemes (piece rate or tournament). We have two dimensions of treatment, the availability of ChatGPT and the suitability of tasks, that constitutes four conditions and it aims to explain the overall effects of ChatGPT on the labor market. This is because different levels of availability and suitability will result in different usage effects or confidence incentives, which will have distinct impacts on willingness to compete and performance. In addition, by comparing the competitive preferences and task performance across groups and genders, we can analyze the impact of ChatGPT usage on gender differences. The results of this study aim to provide substantial recommendations and guidance for future technological development and labor market policies, ensuring that the potential of AIGC tools is fully realised and that technological progress benefits all demographic groups equitably.

Experimental design

In the experimental procedure of this study, we have two types of tasks with different levels of ChatGPT suitability (high suitability and low suitability) and two types of ChatGPT availability (available and unavailable), resulting in a total of four conditions. Participants will be randomly assigned to one of these conditions to follow the steps of the classic competitive preference experiment by Niederle and Vesterlund (2007). Specifically, after completing the first two experimental stages (piece rate and tournament), participants will choose whether to engage in a competitive task in the next stage. The payment scheme will be based either on their future performance (requiring actual work) or their past performance (without requiring actual work). This is designed to measure their competitive preferences while controlling for potential confounding factors.

Design details

Randomization method
Randomization by a computer
Randomization unit
Individual
Clusters
0
Observations
192 students
Arms
48 students under each condition
Primary outcomes
Competition willingness, performances

Cited by

No citation data available for this registration yet.

Keywords