16,000 rated or chosen profiles from 4,000 respondents, in up to 4 tasks of 1 profile varying on 6 attributes.
| Description | 4,000 Japanese adults (Lucid, January 2024) each rated 4 factorial vignettes about a hypothetical 25-year-old rejected for a job or snubbed in a board game club, with target attribute (gender, nationality, sexuality, university), advantaged/disadvantaged group and mechanism randomized; outcome is a 1-6 rating of how likely the scenario is discriminatory. |
|---|---|
| Reference | Igarashi, A., Kano, Y., & Miwa, H. (2025). ChatGPT versus humans in judging discriminatory scenarios: Experimental evidence from a Japanese context. Humanities and Social Sciences Communications, 12, 1776. |
| DOI | 10.1057/s41599-025-06054-6 |
| Licence | CC0 1.0 |
| Source data | https://doi.org/10.7910/DVN/PTON99 |
These notes describe the source data, which the IRW reproduces as released. All notes are listed on the Data Notes page.
| Respondents | 4,000 |
|---|---|
| Rows (one per respondent, task and profile) | 16,000 |
| Tasks per respondent (most) | 4 |
| Profiles per task | 1 |
| Attributes | 6 |
| Topic | social identity interpersonal |
|---|---|
| A profile represents | scenario |
| Country | JP |
| Language shown | ja |
| Language of stored levels | en |
| Randomization restrictions | yes (stated by the source): each of the four targets (gender, nationality, sexuality, education) appears exactly once per respondent; displayed phrases are fixed functions of target x advantaged/disadvantaged status (e.g. Chinese only in a disadvantaged nationality scenario, partner disclosure only in sexuality scenarios, UTokyo/NEPU only in education scenarios); the private sphere has no mechanism |
| Task order | recorded in the deposit |
| Profile position | recorded in the deposit |
Because the attribute levels were not randomized independently, an estimator has to account for the restriction; a plain difference in means across levels can mislead.
Outcomes:
| rating | How likely is the described scenario to be considered discriminatory? (paraphrase) [rating, 1 (Almost certainly not discriminatory) to 6 (Almost certainly discriminatory)] |
|---|
attr_disclosureattr_genderattr_nationalityattr_reasonattr_sphereattr_universitycov_agecov_educationcov_gendercov_prefectureidprofileratingtasktrial_definitiontrial_mechanismtrial_statustrial_target
The layout (one row per respondent, task and profile; choice/rating outcomes; attr_ columns holding each attribute level as displayed) is defined in the IRW conjoint standard. What the outcomes asked is under Design above.
Or load it directly in R or Python (both need a free Redivis account; the first call opens a browser to sign in):
# R
install.packages("remotes")
remotes::install_github("itemresponsewarehouse/Rpkg")
library(irw)
df <- irw_fetch("igarashi_2025_discrimination_judgement", source = "conj")
# Python
pip install irw
import irw
df = irw.fetch("igarashi_2025_discrimination_judgement", source="conj")
@article{Igarashi_2025, title={ChatGPT versus humans in judging discriminatory scenarios: experimental evidence from a Japanese context}, volume={12}, ISSN={2662-9992}, url={http://dx.doi.org/10.1057/s41599-025-06054-6}, DOI={10.1057/s41599-025-06054-6}, number={1}, journal={Humanities and Social Sciences Communications}, publisher={Springer Science and Business Media LLC}, author={Igarashi, Akira and Kano, Yoshinobu and Miwa, Hirofumi}, year={2025}, month=Nov }
| IRW version | v584 |
|---|---|
| Redivis dataset | irw_conjoint v4.2 |
| Redivis dataset DOI | 10.57761/qkye-q530 |
| Manifest pin for this IRW version | v4.2 |
| Metadata source | irw_meta v44.0 |