Generative AI and Labor Market Discrimination

Last registered on May 13, 2024

Pre-Trial

Trial Information

General Information

Title
Generative AI and Labor Market Discrimination
RCT ID
AEARCTR-0013538
Initial registration date
May 01, 2024

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
May 13, 2024, 11:51 AM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Locations

Region

Primary Investigator

Affiliation
Bocconi University

Other Primary Investigator(s)

PI Affiliation
Bocconi University
PI Affiliation
Bocconi University
PI Affiliation
Bocconi University

Additional Trial Information

Status
Completed
Start date
2024-05-01
End date
2024-05-08
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
We study the use of generative AI to produce and screen the CVs and cover letters of job
applicants. With the rise of Large Language Models (LLMs) – such as OpenaAI’s ChatGPT –
there has been a proliferation of websites offering CV generation services for job seekers, and
CV screening services for employers. At the same time, a growing literature has shown that
language models learn stereotyped and biased representations of men and women (Bordia
and Bowman, 2019; Lucy and Bamman, 2021), which can lead them to produce heavily biased
reference letters, for instance (Wan et al., 2023). Building on this recent evidence, we intend to
run a series of pre-registered online experiments to assess the potential consequences of these
production and screening services of CVs and cover letters in the labor market.
External Link(s)

Registration Citation

Citation
Elass, Kenza et al. 2024. "Generative AI and Labor Market Discrimination." AEA RCT Registry. May 13. https://doi.org/10.1257/rct.13538-1.0
Experimental Details

Interventions

Intervention(s)
We prompt large language models to generate and screen CVs.
Intervention (Hidden)
We prompt large language models to generate and screen CVs.

We consider three scenarios:
- the model is told the job seeker is a woman
- the model is told the job seeker is a man
- the model is not told the job seeker's gender
Intervention Start Date
2024-05-01
Intervention End Date
2024-05-08

Primary Outcomes

Primary Outcomes (end points)
Gender differences in CVs generated
Gender differences in screening rates for CVs
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
We prompt large language models to generate and screen CVs of men and women across various occupations. Our treatment variable is gender.
Experimental Design Details
Randomization Method
Randomization by the computer
Randomization Unit
CV
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
45000 CVs
Sample size: planned number of observations
45000 CVs
Sample size (or number of clusters) by treatment arms
45000 CVs
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
IRB

Institutional Review Boards (IRBs)

IRB Name
IRB Approval Date
IRB Approval Number
Analysis Plan

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Post-Trial

Post Trial Information

Study Withdrawal

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Intervention

Is the intervention completed?
No
Data Collection Complete
Data Publication

Data Publication

Is public data available?
No

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials