Table 1.
Survey questions.
Fig 1.
The Figure shows the geographic distribution of study participants from 306 cities across the US.
The blue circle size corresponds to the number of responses from each location. The data reveals a high concentration of participants in urban areas across the Eastern Seaboard, parts of the Midwest, and the West Coast, particularly in California, with sparser distribution in the central United States and rural regions.
Table 2.
Descriptive statistics.
Fig 2.
Heatmap of correlation coefficients between user perception factors of ChatGPT, with actual usage for healthcare-related queries.
Fig 3.
Bayesian correlation sequential analyses of ChatGPT’s attributes and their influence on decision-making assistance.
A: Correlation between perceived competence of ChatGPT and its assistance in decision-making, indicating very strong evidence for the positive association (BF10 = 86.04); B: Association between perceived transparency of ChatGPT and its aid in decision-making, demonstrating strong evidence for the correlation (BF10 = 26.06); C: Relationship between the perceived benefits outweighing risks when using ChatGPT for decision-making, showing strong evidence for the correlation (BF10 = 11.47); D: Correlation between perceived persuasiveness of ChatGPT and its impact on decision-making, with anecdotal evidence for the association (BF10 = 0.67); E: Correlation between perceived trustworthiness and persuasiveness of ChatGPT, suggesting anecdotal evidence for their combined influence on decision-making assistance (BF10 = 0.54); F: Analysis of the relationship between transparency and trustworthiness in ChatGPT, with extreme evidence supporting a very strong correlation (BF10 = 3690).
Table 3.
Comparative Bayesian analysis of decision-making models (n = 44).
Fig 4.
Bayesian model analysis of user perception of ChatGPT.
A: The left panel displays model ranks based on log posterior odds, highlighting the most influential factors determining the intention to use ChatGPT for healthcare inquiries. The colored squares (non-black spaces) indicate the included covariates for a given model, the null model being purple; B: The right panel presents the marginal inclusion probabilities for each factor.
Table 4.
Posterior analysis of predictors influencing decision-making using Bayesian linear regression (n = 44).
Fig 5.
MCMC output analysis from JAGS.
Panel A: displays autocorrelation plots for three components, showing low correlation across lags for multiple chains, indicative of effective sampling. Panel B: presents bivariate scatter plots and marginal histograms for parameter pairs, reflecting their joint and marginal distributions, essential for understanding parameter interactions and individual uncertainties within the model.
Fig 6.
Heatmap of correlation coefficients between user perception factors of ChatGPT, without actual usage for healthcare-related queries.
Table 5.
Comparative Bayesian analysis of decision-making models in non-healthcare contexts (n = 563).
Fig 7.
Bayesian model analysis of non-healthcare user perception of ChatGPT.
A: The left panel displays model ranks based on log posterior odds, highlighting the most influential factors determining the intention to use ChatGPT for healthcare inquiries. The colored squares (non-black spaces) indicate the included covariates for a given model, the null model being purple; B: The right panel presents the marginal inclusion probabilities for each factor.
Table 6.
Posterior analysis of predictors influencing decision-making using Bayesian linear regression (n = 563).
Fig 8.
MCMC output analysis from JAGS.
Panel A: shows autocorrelation plots for seven metrics, indicating chain independence with low autocorrelation at increased lags. Panel B: presents bivariate scatter plots and marginal histograms for parameter pairs, reflecting their joint and marginal distributions, essential for understanding parameter interactions and individual uncertainties within the model.