Fig 1.
Belief that global warming is happening: Distributional comparison of survey and silicon samples.
Note: “Demo Only” represents GPTs are conditioned solely on demographics and “Demo + Cov” represents GPTs are conditioned on demographics and covariates.
Fig 2.
Global warming cause: Distributional comparison of survey and silicon samples.
Note: “Demo Only” represents GPTs are conditioned solely on demographics and “Demo + Cov” represents GPTs are conditioned on demographics and covariates.
Fig 3.
Global warming worry: Distributional comparison of survey and silicon samples.
Note: “Demo Only” represents GPTs are conditioned solely on demographics and “Demo + Cov” represents GPTs are conditioned on demographics and covariates.
Table 1.
Accuracy and MAF1 of GPTs for global warming belief across sub-populations.
Fig 4.
Cramer’s V correlations in survey vs. GPTs.
Note: Cramer’s V could not be estimated for some models because there was no variation in the outcome (e.g., GPT-3.5 with Demo Only for GW Belief).