3,883 responses from 1,963 respondents to 3 items.
One of 12 tables from the same source; this one measures Attitudes to artificial general intelligence (AGI can be built / should be built / should have human rights; 3 items, -3..+3). The others are listed under Related tables.
| Description | Three attitudes to artificial general intelligence -- whether an AGI can be built, whether it should be, and whether it should have human rights -- rated -3 to +3 by 1,963 US adults across three independent samples (Google Surveys 2021, Prolific 2023, Prolific 2024). |
|---|---|
| Reference | Jones, J. J. (2026). Thinking Machines, Pondering Humans: Public Perception of Artificial Intelligence [replication data]. Harvard Dataverse. |
| Licence | CC BY-NC-SA 4.0 |
| Source data | https://doi.org/10.7910/DVN/GJW7TL |
cov_study and each sample's ids sit in their own block (2021 from 1, 2023 from 100001, 2024 from 200001). The 2021 Google Surveys respondents answered one of the three prompts each; the 2023 and 2024 samples answered all three. The 2024 rows come from the deposit's wide correlates file because its long twin carries no respondent id. cov_completion_time_s is the whole-survey duration in seconds, NOT a per-item response time -- in the 2023 file it is identical across a respondent's three answers, and the 2021 survey is one question long.These notes describe the source data, which the IRW reproduces as released. All notes are listed on the Data Notes page.
| Responses | 3,883 |
|---|---|
| Respondents | 1,963 |
| Items | 3 |
| Response categories | 7 |
| Responses per respondent | 1.978 |
| Responses per item | 1294.333 |
| Density | 0.659 |
| Longitudinal | FALSE |
| age range | Adult (18+) |
|---|---|
| sample | General/non-specific, Internet-based |
| construct type | Opinion/attitude |
| measurement tool | Survey/questionnaire |
| item format | Likert Scale/selected response |
| primary language(s) | eng |
| construct name | Attitudes to artificial general intelligence (AGI can be built / should be built / should have human rights; 3 items, -3..+3) |
These are classifications the IRW records, not the study’s own words. Some were annotated by hand and some written by an automated tagger whose output is published only where it clears a measured accuracy bar; which values came from which is not recorded per table. Coverage and accuracy per column, and what is held back, are on Tag quality.
This table has item text in the IRW: the wording administered to respondents, not just the response codes.
| Instrument | Attitudes to artificial general intelligence (Jones, Thinking Machines, Pondering Humans) |
|---|---|
| Mean words per item | 11 |
| Mean characters per item | 56.667 |
| Mean characters per response | 2 |
| Flesch-Kincaid grade level | 5.864 |
This codebook is reconstructed by the IRW from the IRW data standard and the script that built it, jones_2026_ai_perception.py. It is our best reading, not the source’s own codebook, and it may contain mistakes. Where the two disagree, the source data are the authority. If you find an error, please tell us.
| Column | What it is | Source column |
|---|---|---|
| IRW standard columns | ||
id | IRW standard Identifier for the focal unit being measured — typically a person, but sometimes another entity (e.g., a word in a lexical task). | row, recoded in the build script |
item | IRW standard Item identifier. | Prompt, recoded in the build script |
resp | IRW standard Response value. | made in the build script |
| Respondent covariates | ||
cov_age | Covariate | Age |
cov_age_band | Covariate | Age, recoded in the build script |
cov_completion_time_s | Covariate | RT, recoded in the build script |
cov_generalized_trust | Covariate | Generalized_Trust |
cov_party | Covariate | Party_1 |
cov_party_lean | Covariate | Party_2 |
cov_risk_preference | Covariate | Risk_Preference |
cov_sex | Covariate | Sex |
cov_study | Covariate | made in the build script |
What item codes and covariate values stand for is in the codebook released with the source data.
Respondent uniqueness. 6% of this table’s respondents are alone in their combination of the 4 covariates above — no one else in the table shares it. That is a necessary condition for re-identification and not re-identification itself. The IRW republishes the demographics its sources published and never at finer granularity: see the data standard.
Licence: CC BY-NC-SA 4.0 — non-commercial use only; adaptations must be shared under the same licence.
Or load it directly in R (no account needed with irw 1.4.0 or later) or Python (needs a free Redivis account; the first call opens a browser to sign in):
# R
install.packages("irw")
library(irw)
df <- irw_fetch("jones_2026_agi")
# Python
pip install irw
import irw
df = irw.fetch("jones_2026_agi")
Or in Stata, which reads the same no-account CSV as the button above:
* Stata
net install irw, from("https://raw.githubusercontent.com/itemresponsewarehouse/Stata-pkg/main/")
irw_fetch jones_2026_agi, clear
@misc{https://doi.org/10.7910/dvn/gjw7tl,
doi = {10.7910/DVN/GJW7TL},
url = {https://dataverse.harvard.edu/citation?persistentId=doi:10.7910/DVN/GJW7TL},
author = {Jones, Jason},
keywords = {Social Sciences},
title = {Replication Data for: Thinking Machines, Pondering Humans: Public Perception of Artificial Intelligence},
publisher = {Harvard Dataverse},
year = {2026}
}
| IRW version | v744 |
|---|---|
| Redivis dataset | item_response_warehouse_7 v1.4 |
| Redivis dataset DOI | 10.57761/rg8m-en07 |
| Manifest pin for this IRW version | v1.4 |
| Metadata source | irw_meta v78.0 |