84,800 rated or chosen profiles from 4,240 respondents, in up to 10 tasks of 2 profiles varying on 6 attributes.
One of 2 tables from the same source; this one is the loans table. The others are listed under Related tables.
| Description | Paired conjoint in which 4,240 US adults (Qualtrics panel), in a bank loan scenario, chose between and rated 10 pairs of decision-makers (managers or algorithms, by between-respondent arm) described by loan default rate and White and Minority false positive rates. Outcomes: forced choice and a 1-7 performance rating. |
|---|---|
| Reference | Bansak, K., & Paulson, E. (2024). Public attitudes on performance for algorithmic and human decision-makers [Preprint]. OSF Preprints. |
| DOI | 10.31219/osf.io/pghmx |
| Licence | CC0 1.0 |
| Source data | https://doi.org/10.7910/DVN/1QWTHN |
These notes describe the source data, which the IRW reproduces as released. All notes are listed on the Data Notes page.
| Respondents | 4,240 |
|---|---|
| Rows (one per respondent, task and profile) | 84,800 |
| Tasks per respondent (most) | 10 |
| Profiles per task | 2 |
| Attributes | 6 |
| Tasks with no profile chosen | 0 |
| Topic | technology ai privacy |
|---|---|
| A profile represents | person |
| Country | US |
| Language shown | en |
| Language of stored levels | en |
| Randomization restrictions | yes (stated by the source): Decision-maker type is fixed by the respondent's arm (both human, both algorithm, or one of each in faceoff, with the order randomized between respondents); attr_developer applies only to algorithms ('(not shown)' for human DMs); developer and locale are constant within respondent. Rates: uniform independent integers (halves in loans). |
| 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:
| choice | Which decision-maker would you prefer making decisions? (paraphrase) [choice, forced] |
|---|---|
| rating | On a scale from 1 to 7, where 1 indicates VERY BAD and 7 indicates VERY GOOD, how would you rate the performance of [each of these DMs]? [rating, 1 (VERY BAD) to 7 (VERY GOOD)] |
attr_decision_makerattr_default_rateattr_developerattr_localeattr_minority_fprattr_white_fprchoicecov_agecov_ai_enthucov_ai_worrycov_dem_strcov_duration_seccov_educationcov_fair_algo_firmcov_fair_algo_unicov_fair_humancov_gendercov_hispaniccov_incomecov_party_idcov_party_leancov_priority_fpr_equalcov_priority_fpr_lowcov_priority_rate_lowcov_racecov_ratelow_algo_firmcov_ratelow_algo_unicov_ratelow_humancov_rep_strcov_stateidprofileratingtasktrial_fordertrial_matchup
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("bansak_2024_algorithm_dm_loans", source = "conj")
# Python
pip install irw
import irw
df = irw.fetch("bansak_2024_algorithm_dm_loans", source="conj")
@article{Bansak_2023, title={Public Attitudes on Performance for Algorithmic and Human Decision-Makers}, url={http://dx.doi.org/10.31219/osf.io/pghmx}, DOI={10.31219/osf.io/pghmx}, publisher={Center for Open Science}, author={Bansak, Kirk and Paulson, Elisabeth}, year={2023}, month=Oct }
| IRW version | v584 |
|---|---|
| Redivis dataset | irw_conjoint_2 v1.4 |
| Redivis dataset DOI | 10.57761/hk97-wv30 |
| Manifest pin for this IRW version | v1.4 |
| Metadata source | irw_meta v44.0 |