AI Information and High School Students' College Major Choices: Evidence from a Randomized Information Experiment

How to cite

Shi, Chaoqian; Mao, Hui; Huang, Kaixin. 2026. "AI Information and High School Students' College Major Choices: Evidence from a Randomized Information Experiment." AEA RCT Registry. AEARCTR-0018826. https://doi.org/10.1257/rct.18826-1.0

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

Chaoqian Shi* , Hui Mao , Kaixin Huang

* Primary investigator

Registry ID
AEARCTR-0018826
DOI
https://doi.org/10.1257/rct.18826-1.0
First registered
Last updated
Status
In development
Trial dates
2026-06-15 – 2026-12-31
Countries
China (Shaanxi)
JEL codes
I21, I23, J24, D83, O33
Secondary IDs
N/A
Registry record
http://www.socialscienceregistry.org/trials/18826

Abstract

This study uses a randomized information experiment to examine whether information about artificial intelligence and future labor-market changes affects Chinese high school students' intended college major choices, expectations about employment prospects, perceptions of AI substitution and complementarity, and willingness to seek additional major-choice information. Participants complete a baseline questionnaire, are randomly assigned to read different standardized information materials, and then complete comprehension checks and follow-up outcome measures. The intervention does not provide individualized college-major recommendations and is designed as a low-risk educational information treatment.

Intervention

Participating classrooms will be randomly assigned to one of two standardized information materials. Students in control classrooms will receive general information about college major choice, while students in treatment classrooms will receive information about how artificial intelligence may affect future tasks, occupations, and college major choices. The materials do not provide individualized college application advice.

Experimental design

This study uses a two-arm cluster-randomized information experiment with high school students in China. Forty participating classrooms across two high schools will be randomly assigned to one of two information materials. Students first complete consent and baseline questions, receive the information material assigned to their classroom, complete comprehension checks, and then answer follow-up outcome questions. The study compares outcomes between students in classrooms assigned to the AI information group and students in classrooms assigned to the general college-major information group.

Design details

Randomization method
Randomization will be implemented at the classroom level using a pre-specified computer-based random assignment procedure, preferably stratified by high school. Each participating classroom will be assigned to either the control group or the AI information group before survey implementation.
Randomization unit
Classroom.
Clusters
40 classrooms across two high schools
Observations
Approximately 1,500 high school students
Arms
Approximately 20 classrooms assigned to the control group and 20 classrooms assigned to the AI information group, with approximately 750 students in each arm. The exact number of students may vary depending on classroom size, school implementation, and student participation.
Primary outcomes
Change in first-choice intended college major category; change in top-three intended major categories; expectations about employment prospects and income prospects for intended major fields; perceived AI substitution risk and AI complementarity for intended major fields; willingness to search for additional information about majors, employment, and AI impacts; willingness to use AI tools to assist major-choice information search.
Secondary outcomes
AI career anxiety; AI opportunity beliefs; AI self-efficacy; confidence in major choice; perceived credibility, relevance, and comprehension of the information materials; interest in receiving additional major-choice information or AI prompting templates.

Cited by

No citation data available for this registration yet.

Keywords