Questionnaire Selection

This document contains an analysis of the factor structure of the BAIT 1.0 questionnaire (GAAIS + questions about CGI) used in Makowski (Fake Face, 2023).

Item correlation

Code

data <- df |> 
  select(starts_with("AI_")) |> 
  rename(
         GAAIS_Neg3_Unethical = AI_2_Unethical,
         GAAIS_Neg10_Dangerous = AI_6_Dangerous,
         GAAIS_Pos7_DailyLife = AI_4_DailyLife,
         GAAIS_Pos12_Exciting = AI_8_Exciting,
         GAAIS_Pos14_Applications = AI_9_Applications,
         # BAIT
         BAIT_1_RealisticImages = AI_1_RealisticImages,
         BAIT_2_VideosReal = AI_3_VideosReal,
         BAIT_3_ImitatingReality = AI_5_ImitatingReality,
         BAIT_4_RealisticVideos = AI_7_RealisticVideos,
         BAIT_5_FaceErrors = AI_10_FaceErrors)

correlation::correlation(data) |> 
  summary() |> 
  plot() +
  theme_bw() +
  theme(axis.text.x = element_text(angle=45, hjust=1)) 

GAAIS

• General Attitudes towards Artificial Intelligence Scale (GAAIS; Schepman et al., 2020, 2022)

data_gaais <- select(data, starts_with("GAAIS"))

plot(bayestestR::estimate_density(data_gaais, method="KernSmooth")) + 
  theme_abyss() + 
  labs(y = "", x = "Score")

plot(parameters::n_factors(data_gaais))

parameters::factor_analysis(data_gaais, n = 2, rotation = "varimax", sort=TRUE)

# Rotated loadings from Factor Analysis (varimax-rotation)

Variable                 | MR1  | MR2  | Complexity | Uniqueness
----------------------------------------------------------------
GAAIS_Pos12_Exciting     | 0.82 | 0.20 |    1.12    |    0.28   
GAAIS_Pos7_DailyLife     | 0.81 | 0.17 |    1.09    |    0.31   
GAAIS_Pos14_Applications | 0.74 | 0.14 |    1.07    |    0.44   
GAAIS_Neg3_Unethical     | 0.17 | 0.98 |    1.06    |  4.38e-03 
GAAIS_Neg10_Dangerous    | 0.13 | 0.40 |    1.20    |    0.82   

The 2 latent factors (varimax rotation) accounted for 62.84% of the total variance of the original data (MR1 = 38.45%, MR2 = 24.39%).

BAIT

• Beliefs about Artificial Imaging Technology - BAIT (Makowski, preprint)

data_bait <- select(data, starts_with("BAIT"))

plot(bayestestR::estimate_density(data_bait, method="KernSmooth")) + 
  theme_abyss() + 
  labs(y = "", x = "Score")

plot(parameters::n_factors(data_bait))

parameters::factor_analysis(data_bait, n = 2, rotation = "varimax", sort=TRUE)

# Rotated loadings from Factor Analysis (varimax-rotation)

Variable                | MR1  | MR2  | Complexity | Uniqueness
---------------------------------------------------------------
BAIT_4_RealisticVideos  | 0.80 | 0.18 |    1.10    |    0.34   
BAIT_3_ImitatingReality | 0.62 | 0.21 |    1.21    |    0.57   
BAIT_1_RealisticImages  | 0.61 | 0.09 |    1.04    |    0.62   
BAIT_2_VideosReal       | 0.09 | 0.99 |    1.02    |  4.48e-03 
BAIT_5_FaceErrors       | 0.05 | 0.07 |    1.89    |    0.99   

The 2 latent factors (varimax rotation) accounted for 49.48% of the total variance of the original data (MR1 = 28.03%, MR2 = 21.45%).