FictionEro2 - Affective Impact of Fiction

Data Preparation

library(tidyverse)
library(easystats)
library(patchwork)
library(ggside)
library(glmmTMB)

options(mc.cores = parallel::detectCores(),
        brms.backend = "cmdstanr",
        width = 300)

dfsub <- read.csv("../data/data_participants.csv")

df <- read.csv("../data/data.csv") |> 
  right_join(
    select(dfsub, Participant, Mobile, starts_with(c("Feedback_","BAIT")), COPS_Frequency, SexualActivity,- Feedback_Text),
    by = "Participant"
    ) |> 
    datawizard::rescale(select= c("Arousal", "Enticing", "Valence"), range=c(0, 6), to=c(0,1)) |> 
  datawizard::rescale(select= c("Realness"), range=c(-3,3), to=c(0,1)) |>
  mutate(Condition = case_when(
    Condition == "Fiction" ~ "AI-Generated",
    Condition == "Reality" ~ "Photograph"
  )) |>
  mutate(Condition = fct_relevel(Condition, "Photograph", "AI-Generated"),
         Gender =  fct_relevel(Gender, "Male", "Female"),
         SexualOrientation = fct_relevel(SexualOrientation, "Heterosexual", "Homosexual", "Bisexual"),
         PornFrequency = as.numeric(as.factor(COPS_Frequency)),
         SexualActivity_num = as.numeric(as.factor(SexualActivity)),
         ConditionBelief = case_when(
           Condition == "Photograph" & Realness > 0.5 ~ "True",
           Condition == "AI-Generated" & Realness < 0.5 ~ "True",
           .default = "False",
         )) |>
  rename(AllRealConfidence = "Feedback_AllRealConfidence",
         AllFakeConfidence = "Feedback_AllFakeConfidence",
         Enjoyment = "Feedback_Enjoyment"
         )|>
  mutate(across(starts_with("Feedback_"), as.factor)) |>
  mutate(ConditionBelief = as.factor(ConditionBelief)) |>
  mutate(StimuliType = case_when(
    grepl("couple", Stimulus, ignore.case = TRUE) ~ "Couple",
    TRUE ~ "Individual")) |>
  mutate(StimuliType = fct_relevel(StimuliType, "Individual", "Couple"))

results_table <- function(model, effects="fixed", filter=NULL) {
  if("marginaleffects" %in% class(model)) {
    model |> 
      parameters::parameters() |> 
      as.data.frame() |> 
      select(-Parameter, -SE, -S, z=Statistic) |> 
      insight::format_table() |> 
      parameters::display()
  } else {
    display(parameters::parameters(model, effects=effects, keep=filter))
  }
}

df$BAIT_Videos <- (df$BAIT_VideosRealistic + (1 - df$BAIT_VideosIssues)) / 2

df$BAIT_Visual = (df$BAIT_ImagesRealistic + (1 - df$BAIT_ImagesIssues) + df$BAIT_ImitatingReality + df$BAIT_EnvironmentReal) / 4

df$BAIT_Text = (df$BAIT_TextRealistic + (1 - df$BAIT_TextIssues)) / 2

df$BAIT_Negative <- (df$BAIT_Worry + df$BAIT_Dangerous)/2

df$BAIT_Positive <- (df$BAIT_Exciting + df$BAIT_Benefit)/2

Visualisation of Key Variables

This section is about selecting the best model. Notably, all the stimuli presented are of an erotic nature. Note that variables were rescaled to [0,1] when fitting the models and that the stimuli presented to participants was assigned according to their self-rated sexuality, making all stimuli relevant to participants.

To summarise, women rate stimuli on average significantly lower in arousal, enticiment and valence than men. There was no difference between heterosexual, homosexual, and bisexual individuals within each gender on subjective ratings.

Arousal

In general, males exhibited greater variability in their arousal ratings, while females tended to provide more lower arousal ratings.

The pattern of arousal ratings for images that participants believed to be either AI-generated or photos (i.e., initially labeled as photos and later rated as photos, or initially labeled as AI and later rated as AI), as well as for images where participants disagreed with the original labels, was generally similar within genders. However, between genders, women tended to provide more low arousal ratings than men.

arousal <- df |> 
  ggplot(aes(x=Arousal, fill = Gender)) +
  geom_bar() +
  scale_y_sqrt(expand = c(0, 0)) +
  scale_x_continuous(
    breaks = c(0, 1),
    labels = c("Not at all","Very much")) +
  labs(title = "Arousal", ) +
  facet_wrap(Gender~SexualOrientation) +
   scale_fill_manual(values=c("Male"="#E91E63", "Female"="#FF9800"), guide="none") +
  theme_abyss() +
  theme(plot.title = element_text(hjust = 0.5, face="bold"))

arousal

df |> 
  mutate(ConditionBelief = as.factor(ConditionBelief)) |> 
  mutate(ConditionBelief = case_when(
    ConditionBelief == "False" ~ "Not Believed",
    ConditionBelief == "True" ~ "Believed",)) |>
  ggplot(aes(x=Arousal, fill=Gender)) +
  geom_bar(position = "identity") +
   scale_x_continuous(
    breaks = c(0, 1),
    labels = c("Not at all","Very much")) +
  facet_grid(Gender~ConditionBelief, scales="free_y") +
  scale_color_manual(values=c("red", "white"), guide="none") +
  scale_fill_manual(values=c("Male"="#E91E63", "Female"="#FF9800"), guide="none") +
  # scale_y_sqrt(expand = c(0, 0)) +
  labs(title = "Arousal Ratings for Images Based on Belief in Original Labels") +
  theme_abyss() +
  theme(plot.title = element_text(hjust = 0.5, face="bold"))

# Get a sense of Random Factors
m1 <- glmmTMB::glmmTMB(Arousal ~ Gender + (1|Participant), data=df)
m2 <- glmmTMB::glmmTMB(Arousal ~ Gender + (1|Participant) + (1|Stimulus ), data=df)
m3 <- glmmTMB::glmmTMB(Arousal ~ Gender + (1|Participant) + (1+Gender|Stimulus ), data=df)
m4 <- glmmTMB::glmmTMB(Arousal ~ Gender/SexualOrientation + (1|Participant), data=df)
m5 <- glmmTMB::glmmTMB(Arousal ~ Gender/SexualOrientation + (1|Participant) + (1|Stimulus ), data=df)
m6 <- glmmTMB::glmmTMB(Arousal ~ Gender/SexualOrientation + (1|Participant) + (1+Gender|Stimulus ), data=df)


test_bf(m1,m2, m3, m4, m5,m6)
compare_performance(m1, m2, m3, m4, m5, m6)

# Model m3, which includes random intercepts for Participants and both random intercepts and slopes for Gender at the Stimulus level, seems to be the best

This model evaluate ths the effects of Gender and Sexual Orientation on Arousal, accounting for random variability due to participants (i.e., random effects). The analysis revealed that on average women rate erotic stimuli significantly lower in arousal than men. There was no difference between heterosexual, homosexual, and bisexual individuals within each gender on arousal ratings.

m_a <- glmmTMB::glmmTMB(Arousal ~ Gender/ SexualOrientation + (1|Participant),
                      data=df, family=glmmTMB::ordbeta(),
                      control = glmmTMB::glmmTMBControl(parallel = 8))
                      # control = glmmTMB::glmmTMBControl(optimizer=stats::optim, # nlminb, optim
                      #                                   optArgs = list(method = "BFGS"),  #Nelder-Mead, BFGS, CG
                      #                                   parallel = 8))

results_table(m_a)

Fixed Effects

Parameter	Coefficient	SE	95% CI	z	p
(Intercept)	0.08	0.07	(-0.07, 0.22)	1.05	0.295
Gender (Female)	-0.57	0.11	(-0.78, -0.36)	-5.34	< .001
Gender (Male) × SexualOrientationHomosexual	-0.08	0.29	(-0.65, 0.49)	-0.27	0.787
Gender (Female) × SexualOrientationHomosexual	0.16	0.49	(-0.81, 1.13)	0.32	0.746
Gender (Male) × SexualOrientationBisexual	-0.04	0.27	(-0.57, 0.49)	-0.16	0.872
Gender (Female) × SexualOrientationBisexual	0.27	0.21	(-0.15, 0.69)	1.24	0.215

estimate_relation(m_a) |> 
  ggplot(aes(x=Gender, y=Predicted)) +
  geom_pointrange(aes(ymin=CI_low, ymax=CI_high, color = Gender), position=position_dodge(width=0.5)) +
 scale_color_manual(values=c("Male"="#E91E63", "Female"="#FF9800"), guide="none") +  labs(y="Arousal") +
  facet_wrap(~SexualOrientation) +
  theme_abyss()

Enticement

In general, males tended to rate images higher in enticement more often than females.

Enticing ratings for images where participants believed in the original labels, as well as for those where they disagreed with the labels, were similar for within genders, with males providing higher ratings of enticement.

For females, although a similar pattern was observed, they tended to provide more lower enticement ratings, particularly for images where they disagreed with the original labels.

enticiment <- df |> 
  ggplot(aes(x=Enticing, fill = Gender)) +
  geom_bar() +
  scale_y_sqrt(expand = c(0, 0)) +
  scale_x_continuous(
    breaks = c(0, 1),
    labels = c("Not at all","Very much")) +
  labs(title = "Enticement", ) +
  facet_wrap(Gender~SexualOrientation) +
  scale_fill_manual(values=c("Male"="#E91E63", "Female"="#FF9800"), guide="none") +
  theme_abyss() +
  theme(plot.title = element_text(hjust = 0.5, face="bold"))

enticiment

df |> 
  mutate(ConditionBelief = as.factor(ConditionBelief)) |> 
  mutate(ConditionBelief = case_when(
    ConditionBelief == "False" ~ "Not Believed",
    ConditionBelief == "True" ~ "Believed",)) |>
  ggplot(aes(x=Enticing, fill=Gender)) +
  geom_bar(position = "identity", binwidth = 0.5) +
  scale_x_continuous(
    breaks = c(0, 1),
    labels = c("Not at all","Very much")) +
  facet_grid(Gender~ConditionBelief, scales="free_y") +
  scale_fill_manual(values=c("Female"="#E91E63", "Male"="#FF9800"), guide="none") +
  # scale_y_sqrt(expand = c(0, 0)) +
  labs(title = "Enticement Ratings for Images Based on Belief in Original Labels") +
  theme_abyss() +
  theme(plot.title = element_text(hjust = 0.5, face="bold"))

Warning in geom_bar(position = "identity", binwidth = 0.5): Ignoring unknown parameters: `binwidth`

This model evaluate ths the effects of Gender on Enticement, accounting for random variability due to participants (i.e., random effects). The analysis revealed that on average women rate erotic stimuli significantly lower in enticiment than men. There was no difference between heterosexual, homosexual, and bisexual individuals within each gender on enticement ratings.

m_e <- glmmTMB::glmmTMB(Enticing ~ Gender/SexualOrientation + (1|Participant),
                      data=df,family=glmmTMB::ordbeta(),
                      control = glmmTMB::glmmTMBControl(parallel = 8))
                      # control = glmmTMB::glmmTMBControl(optimizer=stats::optim, # nlminb, optim
                      #                                   optArgs = list(method = "BFGS"),  #Nelder-Mead, BFGS, CG
                      #                                   parallel = 8))

results_table(m_e)

Fixed Effects

Parameter	Coefficient	SE	95% CI	z	p
(Intercept)	0.26	0.06	(0.14, 0.38)	4.23	< .001
Gender (Female)	-0.42	0.09	(-0.59, -0.25)	-4.73	< .001
Gender (Male) × SexualOrientationHomosexual	0.06	0.24	(-0.42, 0.54)	0.25	0.801
Gender (Female) × SexualOrientationHomosexual	0.08	0.41	(-0.72, 0.89)	0.20	0.842
Gender (Male) × SexualOrientationBisexual	0.12	0.23	(-0.33, 0.56)	0.51	0.609
Gender (Female) × SexualOrientationBisexual	0.14	0.18	(-0.21, 0.49)	0.78	0.437

estimate_relation(m_e) |> 
  ggplot(aes(x=Gender, y=Predicted)) +
  geom_pointrange(aes(ymin=CI_low, ymax=CI_high, color = Gender), position=position_dodge(width=0.5)) +
 scale_color_manual(values=c("Male"="#E91E63", "Female"="#FF9800"), guide="none") + 
  facet_wrap(~SexualOrientation) +
  labs(y="Enticement") +
  theme_abyss()

Valence

In general, males tended to assign higher valence ratings to images more frequently. Females, on the other hand, showed higher frequencies of ratings in the middle and lower ranges of the valence scale.

The pattern of valence ratings for images, whether believed or not believed, was similar within genders, with ratings for believed images closely resembling those for non-believed images.

valence <- df |> 
  ggplot(aes(x=Valence, fill = Gender)) +
  geom_bar() +
  scale_y_sqrt(expand = c(0, 0)) +
  scale_x_continuous(
    breaks = c(0, 1),
    labels = c("Unplesant","Pleasant")) +
  scale_fill_manual(values=c("Male"="#E91E63", "Female"="#FF9800"), guide="none") +
  labs(title = "Valence", ) +
    facet_wrap(Gender~SexualOrientation) +
  theme_abyss() +
  theme(plot.title = element_text(hjust = 0.5, face="bold"))

valence

df |> 
  mutate(ConditionBelief = as.factor(ConditionBelief)) |> 
  mutate(ConditionBelief = case_when(
    ConditionBelief == "False" ~ "Not Believed",
    ConditionBelief == "True" ~ "Believed",)) |>
  ggplot(aes(x=Valence, fill=Gender)) +
  geom_bar(position = "identity", binwidth = 0.5) +
  scale_x_continuous(
    breaks = c(0, 1),
    labels = c("Unplesant","Pleasant")) +
  facet_grid(Gender~ConditionBelief, scales="free_y") +
  scale_fill_manual(values=c("Female"="#E91E63", "Male"="#FF9800"), guide="none") +
  # scale_y_sqrt(expand = c(0, 0)) +
  labs(title = "Valence to Ratings for Images Based on Belief in Original Labels") +
  theme_abyss() +
  theme(plot.title = element_text(hjust = 0.5, face="bold"))

Warning in geom_bar(position = "identity", binwidth = 0.5): Ignoring unknown parameters: `binwidth`

This model evaluate ths the effects of Gender on valence, accounting for random variability due to participants. The analysis revealed that on average women rate erotic stimuli significantly lower in valence than men. There was no difference between heterosexual, homosexual, and bisexual individuals within each gender on arousal ratings.

m_v <- glmmTMB::glmmTMB(Valence ~ Gender/SexualOrientation+ (1|Participant), family=glmmTMB::ordbeta(),
                      data=df,
                      control = glmmTMB::glmmTMBControl(parallel = 8))
                      # control = glmmTMB::glmmTMBControl(optimizer=stats::optim, # nlminb, optim
                      #                                   optArgs = list(method = "BFGS"),  #Nelder-Mead, BFGS, CG
                      #                                   parallel = 8))

results_table(m_v)

Fixed Effects

Parameter	Coefficient	SE	95% CI	z	p
(Intercept)	0.38	0.06	(0.27, 0.49)	6.84	< .001
Gender (Female)	-0.33	0.08	(-0.49, -0.17)	-4.07	< .001
Gender (Male) × SexualOrientationHomosexual	-0.08	0.22	(-0.51, 0.35)	-0.37	0.711
Gender (Female) × SexualOrientationHomosexual	0.29	0.37	(-0.44, 1.02)	0.77	0.441
Gender (Male) × SexualOrientationBisexual	0.22	0.21	(-0.19, 0.62)	1.06	0.290
Gender (Female) × SexualOrientationBisexual	0.20	0.16	(-0.12, 0.52)	1.25	0.210

estimate_relation(m_v) |> 
  ggplot(aes(x=Gender, y=Predicted)) +
  geom_pointrange(aes(ymin=CI_low, ymax=CI_high, color = Gender), position=position_dodge(width=0.5)) +
 scale_color_manual(values=c("Male"="#E91E63", "Female"="#FF9800"), guide="none") +  labs(y="Valence") +
  facet_wrap(~SexualOrientation) +
  theme_abyss()

df <- df |>
  filter(SexualOrientation == "Heterosexual")

Due to the low number of participants self-described as Bisexual (N= 19) and Homosexual (N= 8) the main analysis will focus solely on Heterosexual individuals.

Exploratory analysis will then evaluate the effect of sexual orientation on ratings.

Arousal

Similar to study1 arousal ratings for some “AI-generated” are lower then when they are presented as real photos.

plot_descriptive <- function(df, target="Arousal") {
  # By item
  order <- df |>
    mutate(.Item_Order = paste0(Gender, "_", Stimulus)) |>
    summarize(mean = mean(.data[[target]], na.rm=TRUE), .by=c("Condition", ".Item_Order")) |> 
    arrange(.Item_Order, desc(Condition)) |> 
    mutate(diff = sign(mean - lag(mean)),
           Relevance = str_extract(.Item_Order, "^[^_]+"),
           .Item_Order = str_extract(.Item_Order, "(?<=_).+")) |>
    filter(Condition == "Photograph") |> 
    arrange(diff, mean)
    
  p1 <- df |>
    mutate(.Item_Order = paste0(Gender, "_", Stimulus)) |>
    group_by(.Item_Order, Condition) |>
    ggdist::mean_qi(.data[[target]], .width=0.95) |>
    mutate(Gender = fct_rev(str_extract(.Item_Order, "^[^_]+"))) |>
    left_join(order[c(".Item_Order", "diff")], by=".Item_Order") |>
mutate(.Item_Order = fct_relevel(.Item_Order, intersect(order$.Item_Order, unique(.Item_Order)))) |>
    # mutate(.average = mean(.data[[target]]), .by="Item") |>
    # mutate(Item = fct_reorder(Item, .average)) |>
    ggplot(aes(y = .Item_Order, x=.data[[target]])) +
    geom_line(aes(color=as.factor(diff)), linewidth=1) +
    geom_point(aes(color=Condition), size=2) +
      
    # ggdist::stat_slabinterval(aes(fill=Condition, color=Condition),
    #                           .width=0.95, normalize="xy", slab_alpha=0.5) +
    # ggdist::stat_histinterval(aes(fill=Condition, color=Condition),
    #                           .width=0.95, normalize="xy", slab_alpha=0.5) +
    scale_color_manual(values=c("AI-Generated"="#2196F3", "Photograph"="#F44336", "1"="#F44336", "-1"="#2196F3"),
                       breaks = c('AI-Generated', 'Photograph')) +
    # scale_fill_manual(values=c("AI-Generated"="#2196F3", "Photograph"="#F44336")) +
    scale_y_discrete(labels = function(x) str_extract(x, "(?<=_).+")) +
    theme_minimal() +
    theme(panel.grid.minor.x = element_blank(),
          # panel.grid.major.x = element_blank(),
          panel.border = element_blank(),
          axis.line.x = element_blank()) +
    labs(title = paste0(target, " by Item"), color="Condition", y="Items") +
    facet_wrap(~Gender, scales="free_y")

  # By participant
  p2 <- df |>
    summarize(.average = mean(.data[[target]]),
              .by=c("Gender", "Participant", "Condition")) |>
    pivot_wider(names_from = Condition, values_from = .average) |>
    mutate(.diff = Photograph - `AI-Generated`) |>
    mutate(Participant = fct_reorder(Participant, .diff)) |>
    ggplot(aes(x = .diff, y=Participant)) +
    geom_bar(aes(fill=.diff), stat = "identity")  +
    scale_fill_gradient2(guide="none", low="red", high="red", mid="blue") +
    labs(title = paste0(target, " by Participant")) +
    facet_wrap(~Gender, ncol=2, scales="free_y") +
    theme_minimal() +
    theme(panel.grid.major.y = element_blank(),
          panel.border = element_blank(),
          axis.line.x = element_blank(),
          axis.text.y = element_blank()) +
    labs(title = paste0(target, " by Participant"), x="Difference (Photograph - AI-Generated)", y="Participants")

  p1 / p2 + plot_layout(heights = c(2, 1))
}

# remove other 
arousal_desc <- df |> 
  filter(Gender != "Other") |> 
  plot_descriptive("Arousal")

arousal_desc

Whole Sample

This model examines the effect of Gender Sexual Orientation, stimuli type, and Condition on Arousal, accounting for random variability due to participants and items (i.e., random intercepts) and condition as random slope for participants.

Women rate photograph of individuals lower than males. Men show less arousal toward couple images (vs individual ones), regardless of condition. Women show greater arousal toward couple images than individual ones. Men show less arousal for AI-generated individual images compared to photographic individual images.

Looking at average slopes:

1. Males rate AI-generated images ~0.014 lower in arousal (on average) than photos — significant.
2. Females also rate AI images lower, but not significantly.
3. Males rate couple images significantly lower in arousal than individual ones.
4. Females rate couple images significantly higher in arousal than individual ones.

m_arousal <- glmmTMB::glmmTMB(Arousal ~ Gender/ StimuliType/ Condition + 
                                (Condition| Participant) + (1|Stimulus), family=glmmTMB::ordbeta(),
                              data=df, control = glmmTMB::glmmTMBControl(parallel = 8))

results_table(m_arousal)

Fixed Effects

Parameter	Coefficient	SE	95% CI	z	p
(Intercept)	0.26	0.12	(0.03, 0.50)	2.17	0.030
Gender (Female)	-1.17	0.17	(-1.52, -0.83)	-6.72	< .001
Gender (Male) × StimuliTypeCouple	-0.39	0.13	(-0.64, -0.14)	-3.02	0.003
Gender (Female) × StimuliTypeCouple	0.60	0.13	(0.34, 0.85)	4.59	< .001
Gender (Male) × StimuliTypeIndividual × ConditionAI-Generated	-0.08	0.04	(-0.15, -0.01)	-2.34	0.019
Gender (Female) × StimuliTypeIndividual × ConditionAI-Generated	0.01	0.05	(-0.08, 0.10)	0.23	0.815
Gender (Male) × StimuliTypeCouple × ConditionAI-Generated	-0.06	0.04	(-0.14, 9.21e-03)	-1.72	0.086
Gender (Female) × StimuliTypeCouple × ConditionAI-Generated	-0.05	0.04	(-0.13, 0.03)	-1.12	0.261

marginaleffects::avg_slopes(m_arousal, variables=c("Condition", "StimuliType"), by="Gender") |>
  results_table()

Comparison	Gender	Coefficient	z	p	95% CI
AI-Generated - Photograph	Male	-0.02	-2.71	0.007	[-0.03, 0.00]
AI-Generated - Photograph	Female	-2.19e-03	-0.34	0.735	[-0.01, 0.01]
Couple - Individual	Male	-0.08	-3.01	0.003	[-0.14, -0.03]
Couple - Individual	Female	0.11	4.50	< .001	[ 0.06, 0.16]

Moderators

This model explores whether various factors such as attitudes towards AI, knowledge about AI, frequency of porn consumption, and sexual activity moderate the relationship between gender, image relevance (specifically AI-generated versus photographs), and arousal ratings.

The analysis revealed a significant interaction effect only for positive attitudes towards AI among females. Specifically, females with higher positive attitudes towards AI rated AI-generated images as more arousing, relative to when they viewed images labelled as photographs.

No other moderators showed statistically significant effects on arousal across the interactions of gender, condition (AI-generated vs Photograph), and the respective moderators.

We also examined whether feedback moderates the relationship between gender and condition on arousal. Findings suggest that feedback moderates this relationship only for men. Specifically, males who reported finding AI-generated images as less and more arousing had lower and higher, respectively, arousal ratings for those images.

Lastly, the interaction between sexual activity and condition on Arousal levels. The three-way interaction between gender, condition (AI-Generated), and sexual activity did not reach statistical significance for either males or females, suggesting that the relationship between the AI-Generated condition and arousal ratings does not significantly vary with sexual activity levels for either gender.

test_moderator <- function(df, outcome="Arousal", moderator="BAIT_Visual") {
  f <- paste0(outcome, " ~ Gender / (Condition * ", moderator, ") + (Condition | Participant) + (1|Stimulus)")
  r <- glmmTMB::glmmTMB(as.formula(f),
                      data = df,
                      family=glmmTMB::ordbeta(),
                      control = glmmTMB::glmmTMBControl(parallel = 8)) |>
    parameters::parameters() |> 
    as.data.frame() |> 
    select(-df_error, -Effects, -Group, -Component, -SE)
  r$Moderator <- moderator
  r$Outcome <- outcome
  insight::format_table(r[c(7, 8), ])
}


moderator_results <- data.frame()
for(m in c("BAIT_Visual", "BAIT_Text", "BAIT_ExpertAI", "BAIT_Positive", "BAIT_Negative", "PornFrequency", "SexualActivity_num")) {
  moderator_results <- rbind(moderator_results,
                             test_moderator(df, "Arousal", m))
}

display(moderator_results)

Parameter	Coefficient	95% CI	z	p	Moderator	Outcome
Gender [Male] × ConditionAI-Generated × BAIT Visual	0.04	[-0.02, 0.10]	1.27	0.205	BAIT_Visual	Arousal
Gender [Female] × ConditionAI-Generated × BAIT Visual	8.80e-03	[-0.06, 0.07]	0.26	0.792	BAIT_Visual	Arousal
Gender [Male] × ConditionAI-Generated × BAIT Text	-0.03	[-0.07, 0.02]	-1.15	0.251	BAIT_Text	Arousal
Gender [Female] × ConditionAI-Generated × BAIT Text	-1.51e-03	[-0.05, 0.05]	-0.06	0.955	BAIT_Text	Arousal
Gender [Male] × ConditionAI-Generated × BAIT ExpertAI	0.02	[-0.01, 0.06]	1.29	0.198	BAIT_ExpertAI	Arousal
Gender [Female] × ConditionAI-Generated × BAIT ExpertAI	0.03	[-0.01, 0.08]	1.56	0.119	BAIT_ExpertAI	Arousal
Gender [Male] × ConditionAI-Generated × BAIT Positive	0.01	[-0.04, 0.06]	0.50	0.620	BAIT_Positive	Arousal
Gender [Female] × ConditionAI-Generated × BAIT Positive	0.07	[ 0.02, 0.12]	2.69	0.007	BAIT_Positive	Arousal
Gender [Male] × ConditionAI-Generated × BAIT Negative	-0.02	[-0.05, 0.01]	-1.07	0.285	BAIT_Negative	Arousal
Gender [Female] × ConditionAI-Generated × BAIT Negative	-0.02	[-0.06, 0.03]	-0.77	0.439	BAIT_Negative	Arousal
Gender [Male] × ConditionAI-Generated × PornFrequency	0.01	[-0.02, 0.05]	0.87	0.384	PornFrequency	Arousal
Gender [Female] × ConditionAI-Generated × PornFrequency	0.01	[-0.03, 0.06]	0.56	0.577	PornFrequency	Arousal
Gender [Male] × ConditionAI-Generated × SexualActivity num	-8.01e-03	[-0.04, 0.03]	-0.47	0.638	SexualActivity_num	Arousal
Gender [Female] × ConditionAI-Generated × SexualActivity num	-0.02	[-0.06, 0.02]	-0.91	0.361	SexualActivity_num	Arousal

feedback_results <- data.frame()
for(m in c("Feedback_NoFacesAttractive", "Feedback_SomeFacesAttractive",
           "Feedback_AIMoreAttractive", "Feedback_AILessAttractive",
           "Feedback_AIMoreArousing", "Feedback_AILessArousing", 
            
           "Feedback_DiffObvious", "Feedback_DiffNone",
           
           "Feedback_LabelsIncorrect", "Feedback_LabelsReversed",
           "Feedback_AllReal", "Feedback_AllFake",
           
           "AllRealConfidence","AllFakeConfidence" ,
           "Enjoyment")) {
  feedback_results <- rbind(feedback_results,
                            test_moderator(df, "Arousal", m))
}

Warning in finalizeTMB(TMBStruc, obj, fit, h, data.tmb.old): Model convergence problem; singular convergence (7). See vignette('troubleshooting'), help('diagnose')
Warning in finalizeTMB(TMBStruc, obj, fit, h, data.tmb.old): Model convergence problem; singular convergence (7). See vignette('troubleshooting'), help('diagnose')

display(feedback_results)

Parameter	Coefficient	95% CI	z	p	Moderator	Outcome
Gender [Male] × ConditionAI-Generated × Feedback NoFacesAttractiveTrue	0.09	[-0.18, 0.35]	0.66	0.511	Feedback_NoFacesAttractive	Arousal
Gender [Female] × ConditionAI-Generated × Feedback NoFacesAttractiveTrue	-0.31	[-0.58, -0.03]	-2.19	0.029	Feedback_NoFacesAttractive	Arousal
Gender [Male] × ConditionAI-Generated × Feedback SomeFacesAttractiveTrue	-0.05	[-0.15, 0.06]	-0.81	0.416	Feedback_SomeFacesAttractive	Arousal
Gender [Female] × ConditionAI-Generated × Feedback SomeFacesAttractiveTrue	0.07	[-0.06, 0.20]	1.10	0.272	Feedback_SomeFacesAttractive	Arousal
Gender [Male] × ConditionAI-Generated × Feedback AIMoreAttractiveTrue	0.14	[ 0.01, 0.28]	2.12	0.034	Feedback_AIMoreAttractive	Arousal
Gender [Female] × ConditionAI-Generated × Feedback AIMoreAttractiveTrue	0.06	[-0.09, 0.20]	0.76	0.448	Feedback_AIMoreAttractive	Arousal
Gender [Male] × ConditionAI-Generated × Feedback AILessAttractiveTrue	-0.18	[-0.36, 0.00]	-1.92	0.054	Feedback_AILessAttractive	Arousal
Gender [Female] × ConditionAI-Generated × Feedback AILessAttractiveTrue	0.03	[-0.18, 0.24]	0.29	0.772	Feedback_AILessAttractive	Arousal
Gender [Male] × ConditionAI-Generated × Feedback AIMoreArousingTrue	0.09	[-0.12, 0.29]	0.83	0.407	Feedback_AIMoreArousing	Arousal
Gender [Female] × ConditionAI-Generated × Feedback AIMoreArousingTrue	-1.92e-03	[-0.35, 0.35]	-0.01	0.991	Feedback_AIMoreArousing	Arousal
Gender [Male] × ConditionAI-Generated × Feedback AILessArousingTrue	0.11	[-0.10, 0.31]	0.99	0.322	Feedback_AILessArousing	Arousal
Gender [Female] × ConditionAI-Generated × Feedback AILessArousingTrue	0.07	[-0.21, 0.36]	0.51	0.608	Feedback_AILessArousing	Arousal
Gender [Male] × ConditionAI-Generated × Feedback DiffObviousTrue	1.08e-03	[-0.23, 0.23]	9.10e-03	0.993	Feedback_DiffObvious	Arousal
Gender [Female] × ConditionAI-Generated × Feedback DiffObviousTrue	-0.19	[-0.49, 0.10]	-1.27	0.203	Feedback_DiffObvious	Arousal
Gender [Male] × ConditionAI-Generated × Feedback DiffNoneTrue	0.01	[-0.10, 0.12]	0.25	0.806	Feedback_DiffNone	Arousal
Gender [Female] × ConditionAI-Generated × Feedback DiffNoneTrue	0.02	[-0.12, 0.15]	0.25	0.804	Feedback_DiffNone	Arousal
Gender [Male] × ConditionAI-Generated × Feedback LabelsIncorrectTrue	-0.06	[-0.19, 0.07]	-0.92	0.359	Feedback_LabelsIncorrect	Arousal
Gender [Female] × ConditionAI-Generated × Feedback LabelsIncorrectTrue	0.11	[-0.03, 0.26]	1.50	0.133	Feedback_LabelsIncorrect	Arousal
Gender [Male] × ConditionAI-Generated × Feedback LabelsReversedTrue	0.20	[-0.05, 0.46]	1.54	0.122	Feedback_LabelsReversed	Arousal
Gender [Female] × ConditionAI-Generated × Feedback LabelsReversedTrue	-0.13	[-0.47, 0.21]	-0.77	0.439	Feedback_LabelsReversed	Arousal
Gender [Male] × ConditionAI-Generated × Feedback AllRealTrue	3.71e-04	[-0.11, 0.12]	6.31e-03	0.995	Feedback_AllReal	Arousal
Gender [Female] × ConditionAI-Generated × Feedback AllRealTrue	0.01	[-0.12, 0.15]	0.22	0.827	Feedback_AllReal	Arousal
Gender [Male] × ConditionAI-Generated × Feedback AllFakeTrue	0.14	[-0.09, 0.36]	1.16	0.244	Feedback_AllFake	Arousal
Gender [Female] × ConditionAI-Generated × Feedback AllFakeTrue	-0.12	[-0.43, 0.18]	-0.81	0.417	Feedback_AllFake	Arousal
Gender [Male] × ConditionAI-Generated × AllRealConfidence	-5.04e-03	[-0.07, 0.06]	-0.14	0.887	AllRealConfidence	Arousal
Gender [Female] × ConditionAI-Generated × AllRealConfidence	0.02	[-0.06, 0.09]	0.41	0.678	AllRealConfidence	Arousal
Gender [Male] × ConditionAI-Generated × AllFakeConfidence	0.04	[-0.23, 0.30]	0.26	0.798	AllFakeConfidence	Arousal
Gender [Female] × ConditionAI-Generated × AllFakeConfidence	-0.10	[-0.48, 0.27]	-0.54	0.592	AllFakeConfidence	Arousal
Gender [Male] × ConditionAI-Generated × Enjoyment	-2.76e-03	[-0.09, 0.09]	-0.06	0.952	Enjoyment	Arousal
Gender [Female] × ConditionAI-Generated × Enjoyment	0.02	[-0.06, 0.09]	0.43	0.664	Enjoyment	Arousal

m <- glmmTMB::glmmTMB(Arousal ~ Gender / (Condition * SexualActivity_num) + (Condition | Participant) + (1|Stimulus),
                      data=filter(df, ConditionBelief == "True"), 
                      family=glmmTMB::ordbeta(),
                      control = glmmTMB::glmmTMBControl(parallel = 8))

results_table(m)

Fixed Effects

Parameter	Coefficient	SE	95% CI	z	p
(Intercept)	-0.20	0.18	(-0.55, 0.14)	-1.15	0.252
Gender (Female)	-0.30	0.26	(-0.81, 0.21)	-1.16	0.247
Gender (Male) × ConditionAI-Generated	-0.29	0.09	(-0.47, -0.12)	-3.35	< .001
Gender (Female) × ConditionAI-Generated	-0.23	0.13	(-0.48, 0.02)	-1.82	0.069
Gender (Male) × SexualActivity num	0.04	0.05	(-0.06, 0.13)	0.78	0.438
Gender (Female) × SexualActivity num	0.05	0.05	(-0.06, 0.16)	0.89	0.373
Gender (Male) × ConditionAI-Generated × SexualActivity num	-3.09e-03	0.03	(-0.05, 0.05)	-0.12	0.905
Gender (Female) × ConditionAI-Generated × SexualActivity num	0.03	0.03	(-0.04, 0.10)	0.90	0.366

sex_act_plot<- estimate_relation(m, length=20) |> 
  ggplot(aes(x=SexualActivity_num, y=Predicted)) +
  geom_ribbon(aes(ymin=CI_low, ymax=CI_high, fill=Condition), alpha=0.2) +
  geom_line(aes(color=Condition)) +
  facet_grid(~Gender)

sex_act_plot

Believed

m <- glmmTMB::glmmTMB(Arousal ~ Gender / StimuliType/Condition + (Condition|Participant) + (Condition|Stimulus),
                        data=filter(df, ConditionBelief == "True"), 
                        family=glmmTMB::ordbeta(),
                        control = glmmTMB::glmmTMBControl(parallel = 8))
  
results_table(m)

Fixed Effects

Parameter	Coefficient	SE	95% CI	z	p
(Intercept)	0.37	0.12	(0.13, 0.61)	3.02	0.002
Gender (Female)	-1.26	0.18	(-1.61, -0.91)	-7.05	< .001
Gender (Male) × StimuliTypeCouple	-0.45	0.13	(-0.71, -0.19)	-3.35	< .001
Gender (Female) × StimuliTypeCouple	0.62	0.14	(0.35, 0.89)	4.45	< .001
Gender (Male) × StimuliTypeIndividual × ConditionAI-Generated	-0.31	0.06	(-0.43, -0.20)	-5.38	< .001
Gender (Female) × StimuliTypeIndividual × ConditionAI-Generated	-0.03	0.08	(-0.18, 0.12)	-0.39	0.696
Gender (Male) × StimuliTypeCouple × ConditionAI-Generated	-0.31	0.06	(-0.44, -0.19)	-4.91	< .001
Gender (Female) × StimuliTypeCouple × ConditionAI-Generated	-0.17	0.07	(-0.31, -0.03)	-2.39	0.017

moderator_believed <- data.frame()
for(m in c("BAIT_Visual", "BAIT_Text", "BAIT_ExpertAI",
             "BAIT_Positive", "BAIT_Negative",
             "PornFrequency", "SexualActivity", 
             "Feedback_DiffObvious", "Feedback_DiffNone",
             "Feedback_AIMoreArousing", "Feedback_AILessArousing", 
             "Feedback_AIMoreAttractive", "Feedback_AILessAttractive",
             "Feedback_LabelsIncorrect", "Feedback_LabelsReversed")) {
    moderator_believed <- rbind(moderator_believed,
                               test_moderator(filter(df, ConditionBelief == "True"), "Arousal", m))
  }
  
  display(filter(moderator_believed, p < .1))

Parameter	Coefficient	95% CI	z	p	Moderator	Outcome
Gender [Female] × ConditionAI-Generated × PornFrequency	0.06	[-0.01, 0.13]	1.65	0.099	PornFrequency	Arousal
Gender [Female] × SexualActivitywithin the last 3 days	0.50	[ 0.02, 0.98]	2.03	0.042	SexualActivity	Arousal
Gender [Male] × ConditionAI-Generated × Feedback AIMoreArousingTrue	0.25	[-0.04, 0.54]	1.72	0.086	Feedback_AIMoreArousing	Arousal

Enticement

Enticment ratings, for some images, differ by whether the images were preceeded by ‘photograph’ or by ‘AI-generated’.

enticement_desc <- plot_descriptive(df, "Enticing")
enticement_desc

Whole Sample

This model examines how Gender, Stimuli Type and Condition influence enticement ratings, treating participant and items as random intercepts.

Women rated images as less enticing than men overall. They rated couple images as more enticing than individual images. Men, on the other hand, rated couple images as less enticing. Men also rated AI-generated images as significantly less enticing than photographs, both for images of individual and couple stimuli. In contrast, women did not significantly differentiate between AI and photographic images in terms of enticement.

Looking at the average slopes, both genders rate Ai-generated conditions lower in enticement, although this effect is only significant for men.

m_enticement <- glmmTMB::glmmTMB(Enticing ~ Gender / StimuliType/ Condition + 
                                   (Condition | Participant) + (1|Stimulus),
                      data=df, family=glmmTMB::ordbeta(),
                      control = glmmTMB::glmmTMBControl(parallel = 8))



results_table(m_enticement)

Fixed Effects

Parameter	Coefficient	SE	95% CI	z	p
(Intercept)	0.50	0.10	(0.31, 0.70)	5.03	< .001
Gender (Female)	-0.98	0.14	(-1.26, -0.70)	-6.83	< .001
Gender (Male) × StimuliTypeCouple	-0.38	0.11	(-0.59, -0.17)	-3.61	< .001
Gender (Female) × StimuliTypeCouple	0.53	0.11	(0.32, 0.74)	4.95	< .001
Gender (Male) × StimuliTypeIndividual × ConditionAI-Generated	-0.10	0.04	(-0.17, -0.03)	-2.83	0.005
Gender (Female) × StimuliTypeIndividual × ConditionAI-Generated	0.05	0.04	(-0.04, 0.13)	1.12	0.262
Gender (Male) × StimuliTypeCouple × ConditionAI-Generated	-0.08	0.04	(-0.15, -0.01)	-2.28	0.022
Gender (Female) × StimuliTypeCouple × ConditionAI-Generated	-0.04	0.04	(-0.12, 0.03)	-1.14	0.253

marginaleffects::avg_slopes(m_enticement, variables="Condition", by="Gender") |>
  results_table()

Comparison	Gender	Coefficient	z	p	95% CI
AI-Generated - Photograph	Male	-0.02	-3.33	< .001	[-0.03, -0.01]
AI-Generated - Photograph	Female	1.50e-04	0.02	0.982	[-0.01, 0.01]

Moderators

We examined whether attitudes towards AI, porn frequency, and sexual activity moderated the relationship between the predictors (sex, relevance, condition) and enticement ratings. The results showed that enticing ratings for images labeled AI-generated were moderated by the feedback reporting attitudes towards AI, but this was only significant for women. Specifically, women who reported having more positive attitudes towards AI, rated these images higher in arousal compared to images labeled as photographs.

Looking at feedback as a moderator, the results showed that enticement ratings for images labelled as AI-Generated were moderated by the feedback that AI is less attractive for males, and that difference between the AI-Generated images and real photos is obvious for females. Specifically,

Lastly, the interaction between sexual activity and condition on enticement, suggests that enticement ratings do not significantly differ with sexual activity levels for either gender.

for(m in c("BAIT_Visual", "BAIT_Text", "BAIT_ExpertAI",
           "BAIT_Positive", "BAIT_Negative",
           "PornFrequency", "SexualActivity")) {
  moderator_results <- rbind(moderator_results,
                             test_moderator(df, "Enticing", m))
}

display(filter(moderator_results, Outcome == "Enticing"))

Parameter	Coefficient	95% CI	z	p	Moderator	Outcome
Gender [Male] × ConditionAI-Generated × BAIT Visual	0.03	[-0.03, 0.09]	0.92	0.355	BAIT_Visual	Enticing
Gender [Female] × ConditionAI-Generated × BAIT Visual	2.94e-03	[-0.05, 0.06]	0.10	0.919	BAIT_Visual	Enticing
Gender [Male] × ConditionAI-Generated × BAIT Text	-0.01	[-0.06, 0.03]	-0.47	0.639	BAIT_Text	Enticing
Gender [Female] × ConditionAI-Generated × BAIT Text	-8.19e-04	[-0.05, 0.05]	-0.03	0.973	BAIT_Text	Enticing
Gender [Male] × ConditionAI-Generated × BAIT ExpertAI	0.02	[-0.01, 0.05]	1.13	0.257	BAIT_ExpertAI	Enticing
Gender [Female] × ConditionAI-Generated × BAIT ExpertAI	2.02e-03	[-0.04, 0.04]	0.10	0.921	BAIT_ExpertAI	Enticing
Gender [Male] × ConditionAI-Generated × BAIT Positive	0.01	[-0.04, 0.06]	0.49	0.624	BAIT_Positive	Enticing
Gender [Female] × ConditionAI-Generated × BAIT Positive	0.05	[ 0.00, 0.09]	1.96	0.050	BAIT_Positive	Enticing
Gender [Male] × ConditionAI-Generated × BAIT Negative	-0.04	[-0.07, -0.01]	-2.38	0.017	BAIT_Negative	Enticing
Gender [Female] × ConditionAI-Generated × BAIT Negative	1.45e-03	[-0.04, 0.04]	0.07	0.943	BAIT_Negative	Enticing
Gender [Male] × ConditionAI-Generated × PornFrequency	-3.50e-03	[-0.04, 0.03]	-0.21	0.835	PornFrequency	Enticing
Gender [Female] × ConditionAI-Generated × PornFrequency	-0.01	[-0.06, 0.03]	-0.62	0.533	PornFrequency	Enticing
Gender [Male] × SexualActivitywithin the last 3 days	-2.15e-03	[-0.33, 0.33]	-0.01	0.990	SexualActivity	Enticing
Gender [Female] × SexualActivitywithin the last 3 days	0.51	[ 0.11, 0.92]	2.49	0.013	SexualActivity	Enticing

for(m in c("Feedback_NoFacesAttractive", "Feedback_SomeFacesAttractive",
           "Feedback_AIMoreAttractive", "Feedback_AILessAttractive",
           "Feedback_AIMoreArousing", "Feedback_AILessArousing", 
            
           "Feedback_DiffObvious", "Feedback_DiffNone",
           
           "Feedback_LabelsIncorrect", "Feedback_LabelsReversed",
           "Feedback_AllReal", "Feedback_AllFake",
           
           "AllRealConfidence","AllFakeConfidence" ,
           "Enjoyment")) {
  feedback_results <- rbind(feedback_results,
                            test_moderator(df, "Enticing", m))
}

Warning in finalizeTMB(TMBStruc, obj, fit, h, data.tmb.old): Model convergence problem; singular convergence (7). See vignette('troubleshooting'), help('diagnose')

display(filter(feedback_results, Outcome == "Enticing"))

Parameter	Coefficient	95% CI	z	p	Moderator	Outcome
Gender [Male] × ConditionAI-Generated × Feedback NoFacesAttractiveTrue	0.05	[-0.22, 0.31]	0.36	0.716	Feedback_NoFacesAttractive	Enticing
Gender [Female] × ConditionAI-Generated × Feedback NoFacesAttractiveTrue	-0.09	[-0.33, 0.16]	-0.71	0.478	Feedback_NoFacesAttractive	Enticing
Gender [Male] × ConditionAI-Generated × Feedback SomeFacesAttractiveTrue	-8.85e-03	[-0.12, 0.10]	-0.16	0.874	Feedback_SomeFacesAttractive	Enticing
Gender [Female] × ConditionAI-Generated × Feedback SomeFacesAttractiveTrue	-0.02	[-0.14, 0.10]	-0.28	0.780	Feedback_SomeFacesAttractive	Enticing
Gender [Male] × ConditionAI-Generated × Feedback AIMoreAttractiveTrue	0.07	[-0.06, 0.21]	1.07	0.285	Feedback_AIMoreAttractive	Enticing
Gender [Female] × ConditionAI-Generated × Feedback AIMoreAttractiveTrue	-8.89e-03	[-0.15, 0.13]	-0.13	0.900	Feedback_AIMoreAttractive	Enticing
Gender [Male] × ConditionAI-Generated × Feedback AILessAttractiveTrue	-0.18	[-0.36, 0.01]	-1.88	0.060	Feedback_AILessAttractive	Enticing
Gender [Female] × ConditionAI-Generated × Feedback AILessAttractiveTrue	3.67e-03	[-0.19, 0.20]	0.04	0.970	Feedback_AILessAttractive	Enticing
Gender [Male] × ConditionAI-Generated × Feedback AIMoreArousingTrue	0.03	[-0.17, 0.23]	0.27	0.784	Feedback_AIMoreArousing	Enticing
Gender [Female] × ConditionAI-Generated × Feedback AIMoreArousingTrue	-0.09	[-0.41, 0.23]	-0.55	0.585	Feedback_AIMoreArousing	Enticing
Gender [Male] × ConditionAI-Generated × Feedback AILessArousingTrue	0.04	[-0.18, 0.25]	0.33	0.739	Feedback_AILessArousing	Enticing
Gender [Female] × ConditionAI-Generated × Feedback AILessArousingTrue	0.17	[-0.07, 0.42]	1.38	0.166	Feedback_AILessArousing	Enticing
Gender [Male] × ConditionAI-Generated × Feedback DiffObviousTrue	-0.14	[-0.37, 0.09]	-1.17	0.241	Feedback_DiffObvious	Enticing
Gender [Female] × ConditionAI-Generated × Feedback DiffObviousTrue	-0.18	[-0.45, 0.09]	-1.29	0.197	Feedback_DiffObvious	Enticing
Gender [Male] × ConditionAI-Generated × Feedback DiffNoneTrue	0.04	[-0.07, 0.15]	0.72	0.470	Feedback_DiffNone	Enticing
Gender [Female] × ConditionAI-Generated × Feedback DiffNoneTrue	5.57e-03	[-0.12, 0.13]	0.09	0.928	Feedback_DiffNone	Enticing
Gender [Male] × ConditionAI-Generated × Feedback LabelsIncorrectTrue	-0.06	[-0.19, 0.07]	-0.84	0.399	Feedback_LabelsIncorrect	Enticing
Gender [Female] × ConditionAI-Generated × Feedback LabelsIncorrectTrue	0.02	[-0.12, 0.16]	0.27	0.784	Feedback_LabelsIncorrect	Enticing
Gender [Male] × ConditionAI-Generated × Feedback LabelsReversedTrue	0.17	[-0.10, 0.43]	1.25	0.212	Feedback_LabelsReversed	Enticing
Gender [Female] × ConditionAI-Generated × Feedback LabelsReversedTrue	-0.14	[-0.46, 0.17]	-0.89	0.375	Feedback_LabelsReversed	Enticing
Gender [Male] × ConditionAI-Generated × Feedback AllRealTrue	0.02	[-0.09, 0.13]	0.35	0.724	Feedback_AllReal	Enticing
Gender [Female] × ConditionAI-Generated × Feedback AllRealTrue	0.08	[-0.04, 0.21]	1.37	0.171	Feedback_AllReal	Enticing
Gender [Male] × ConditionAI-Generated × Feedback AllFakeTrue	0.21	[-0.02, 0.44]	1.79	0.073	Feedback_AllFake	Enticing
Gender [Female] × ConditionAI-Generated × Feedback AllFakeTrue	-3.78e-03	[-0.28, 0.27]	-0.03	0.979	Feedback_AllFake	Enticing
Gender [Male] × ConditionAI-Generated × AllRealConfidence	-0.03	[-0.09, 0.03]	-0.91	0.363	AllRealConfidence	Enticing
Gender [Female] × ConditionAI-Generated × AllRealConfidence	0.01	[-0.06, 0.08]	0.31	0.758	AllRealConfidence	Enticing
Gender [Male] × ConditionAI-Generated × AllFakeConfidence	-5.32e-03	[-0.29, 0.28]	-0.04	0.970	AllFakeConfidence	Enticing
Gender [Female] × ConditionAI-Generated × AllFakeConfidence	0.03	[-0.31, 0.37]	0.16	0.870	AllFakeConfidence	Enticing
Gender [Male] × ConditionAI-Generated × Enjoyment	0.04	[-0.05, 0.13]	0.90	0.369	Enjoyment	Enticing
Gender [Female] × ConditionAI-Generated × Enjoyment	-5.09e-03	[-0.07, 0.06]	-0.16	0.873	Enjoyment	Enticing

m <- glmmTMB::glmmTMB(Enticing ~ Gender / (Condition * SexualActivity_num) + (Condition | Participant) + (1|Stimulus),
                      data=filter(df, ConditionBelief == "True"), 
                      family=glmmTMB::ordbeta(),
                      control = glmmTMB::glmmTMBControl(parallel = 8))

results_table(m)

Fixed Effects

Parameter	Coefficient	SE	95% CI	z	p
(Intercept)	0.15	0.15	(-0.15, 0.44)	0.97	0.334
Gender (Female)	-0.32	0.22	(-0.75, 0.12)	-1.44	0.150
Gender (Male) × ConditionAI-Generated	-0.36	0.10	(-0.55, -0.17)	-3.77	< .001
Gender (Female) × ConditionAI-Generated	-0.12	0.12	(-0.36, 0.12)	-0.99	0.323
Gender (Male) × SexualActivity num	0.02	0.04	(-0.07, 0.10)	0.38	0.707
Gender (Female) × SexualActivity num	0.06	0.05	(-0.03, 0.15)	1.37	0.171
Gender (Male) × ConditionAI-Generated × SexualActivity num	0.01	0.03	(-0.04, 0.07)	0.52	0.604
Gender (Female) × ConditionAI-Generated × SexualActivity num	6.23e-03	0.03	(-0.06, 0.07)	0.19	0.852

sex_act_plot_ent<- estimate_relation(m, length=20) |> 
  ggplot(aes(x=SexualActivity_num, y=Predicted)) +
  geom_ribbon(aes(ymin=CI_low, ymax=CI_high, fill=Condition), alpha=0.2) +
  geom_line(aes(color=Condition)) +
  facet_grid(~Gender)

sex_act_plot_ent

Believed

m <- glmmTMB::glmmTMB(Enticing ~ Gender / Condition + (Condition| Participant) + (1|Stimulus),
                      data=filter(df, ConditionBelief == "True"), 
                      family=glmmTMB::ordbeta(),
                      control = glmmTMB::glmmTMBControl(parallel = 8))

results_table(m)

Fixed Effects

Parameter	Coefficient	SE	95% CI	z	p
(Intercept)	0.19	0.09	(0.02, 0.37)	2.13	0.033
Gender (Female)	-0.15	0.10	(-0.35, 0.05)	-1.50	0.134
Gender (Male) × ConditionAI-Generated	-0.32	0.05	(-0.41, -0.22)	-6.67	< .001
Gender (Female) × ConditionAI-Generated	-0.10	0.05	(-0.20, 7.43e-04)	-1.95	0.052

for(m in c("BAIT_Visual", "BAIT_Text", "BAIT_ExpertAI",
           "BAIT_Positive", "BAIT_Negative", "PornFrequency", "SexualActivity", 
           
           "Feedback_DiffObvious", "Feedback_DiffNone",
           "Feedback_AIMoreArousing", "Feedback_AILessArousing",
           "Feedback_AIMoreAttractive", "Feedback_AILessArousing",
           "Feedback_LabelsIncorrect", "Feedback_LabelsReversed")) {
  moderator_believed <- rbind(moderator_believed,
                             test_moderator(filter(df, ConditionBelief == "True"), "Enticing", m))
}

display(filter(moderator_believed, p < .1))

Parameter	Coefficient	95% CI	z	p	Moderator	Outcome
Gender [Female] × ConditionAI-Generated × PornFrequency	0.06	[-0.01, 0.13]	1.65	0.099	PornFrequency	Arousal
Gender [Female] × SexualActivitywithin the last 3 days	0.50	[ 0.02, 0.98]	2.03	0.042	SexualActivity	Arousal
Gender [Male] × ConditionAI-Generated × Feedback AIMoreArousingTrue	0.25	[-0.04, 0.54]	1.72	0.086	Feedback_AIMoreArousing	Arousal
Gender [Female] × SexualActivitywithin the last 3 days	0.53	[ 0.13, 0.94]	2.56	0.010	SexualActivity	Enticing

Valence

valence_desc <- plot_descriptive(df, "Valence")
valence_desc

Whole Sample

This model examines the effect of Gender, Stimuli Type and Condition on Valence ratings, while accounting for random variability due to participants and items (i.e., random intercepts).

Women rated images overall as lower in valence than men did. They rated couple images as higher in valence than individual ones. Additionally, women rated AI-generated couple images as significantly lower in valence than photographic couple images. Men rated images of couples lower in valence and AI-images of individuals lower in valence comapred to real photos of individuals.

Looking at avergae slopes, both genders rate AI-generated images lower in valence, but this effect is only significant for males as well.

m_valence <- glmmTMB::glmmTMB(Valence ~ Gender / StimuliType/ Condition + 
                                (Condition| Participant) + (1|Stimulus),
                      data=df, family=glmmTMB::ordbeta(),
                      control = glmmTMB::glmmTMBControl(parallel = 8))

results_table(m_valence)

Fixed Effects

Parameter	Coefficient	SE	95% CI	z	p
(Intercept)	0.57	0.08	(0.41, 0.73)	7.05	< .001
Gender (Female)	-0.77	0.12	(-0.99, -0.54)	-6.60	< .001
Gender (Male) × StimuliTypeCouple	-0.27	0.08	(-0.43, -0.12)	-3.39	< .001
Gender (Female) × StimuliTypeCouple	0.48	0.08	(0.32, 0.64)	5.79	< .001
Gender (Male) × StimuliTypeIndividual × ConditionAI-Generated	-0.08	0.03	(-0.15, -0.02)	-2.43	0.015
Gender (Female) × StimuliTypeIndividual × ConditionAI-Generated	0.01	0.04	(-0.06, 0.08)	0.30	0.763
Gender (Male) × StimuliTypeCouple × ConditionAI-Generated	-0.04	0.03	(-0.10, 0.03)	-1.15	0.249
Gender (Female) × StimuliTypeCouple × ConditionAI-Generated	-0.06	0.03	(-0.13, 0.01)	-1.66	0.098

marginaleffects::avg_slopes(m_valence, variables="Condition", by="Gender") |>
  results_table()

Comparison	Gender	Coefficient	z	p	95% CI
AI-Generated - Photograph	Male	-0.01	-2.51	0.012	[-0.02, 0.00]
AI-Generated - Photograph	Female	-6.08e-03	-1.02	0.308	[-0.02, 0.01]

Moderators

Here we look at whether the relationship between gender and condition and valence is moderated by AI-attitudes, porn frequency, and sexual activity. The results showed that enticing ratings for images labeled AI-generated were moderated by the feedback reporting attitudes towards AI, but this was only significant for women. More specifically, females who have more positive and negative attitudes towards AI rated images labelled as AI-Generated significantly higher and lower in valence compared to images labelled as photographs.

Looking at feedback as a moderator, the results show that valence ratings for images labelled AI-Generated were moderted by the feedback that the difference between AI and Photos was obvious, but only for females.

for(m in c("BAIT_Visual", "BAIT_Text", "BAIT_ExpertAI",
           "BAIT_Positive", "BAIT_Negative", "PornFrequency", "SexualActivity_num")) {
  moderator_results <- rbind(moderator_results,
                             test_moderator(df, "Valence", m))
}

display(filter(moderator_results, Outcome == "Valence"))

Parameter	Coefficient	95% CI	z	p	Moderator	Outcome
Gender [Male] × ConditionAI-Generated × BAIT Visual	0.04	[-0.01, 0.09]	1.53	0.127	BAIT_Visual	Valence
Gender [Female] × ConditionAI-Generated × BAIT Visual	8.76e-03	[-0.04, 0.06]	0.35	0.726	BAIT_Visual	Valence
Gender [Male] × ConditionAI-Generated × BAIT Text	-3.72e-04	[-0.04, 0.04]	-0.02	0.985	BAIT_Text	Valence
Gender [Female] × ConditionAI-Generated × BAIT Text	0.02	[-0.02, 0.06]	0.99	0.323	BAIT_Text	Valence
Gender [Male] × ConditionAI-Generated × BAIT ExpertAI	0.04	[ 0.01, 0.07]	2.51	0.012	BAIT_ExpertAI	Valence
Gender [Female] × ConditionAI-Generated × BAIT ExpertAI	0.02	[-0.01, 0.05]	1.11	0.269	BAIT_ExpertAI	Valence
Gender [Male] × ConditionAI-Generated × BAIT Positive	0.02	[-0.02, 0.07]	1.04	0.300	BAIT_Positive	Valence
Gender [Female] × ConditionAI-Generated × BAIT Positive	0.03	[-0.01, 0.07]	1.71	0.087	BAIT_Positive	Valence
Gender [Male] × ConditionAI-Generated × BAIT Negative	-0.03	[-0.06, 0.00]	-2.30	0.022	BAIT_Negative	Valence
Gender [Female] × ConditionAI-Generated × BAIT Negative	-0.04	[-0.07, 0.00]	-2.24	0.025	BAIT_Negative	Valence
Gender [Male] × ConditionAI-Generated × PornFrequency	-1.16e-03	[-0.03, 0.03]	-0.08	0.937	PornFrequency	Valence
Gender [Female] × ConditionAI-Generated × PornFrequency	-2.57e-03	[-0.04, 0.03]	-0.14	0.888	PornFrequency	Valence
Gender [Male] × ConditionAI-Generated × SexualActivity num	-0.02	[-0.05, 0.01]	-1.17	0.243	SexualActivity_num	Valence
Gender [Female] × ConditionAI-Generated × SexualActivity num	-0.01	[-0.04, 0.02]	-0.70	0.481	SexualActivity_num	Valence

for(m in c("Feedback_NoFacesAttractive", "Feedback_SomeFacesAttractive",
           "Feedback_AIMoreAttractive", "Feedback_AILessAttractive",
           "Feedback_AIMoreArousing", "Feedback_AILessArousing", 
            
           "Feedback_DiffObvious", "Feedback_DiffNone",
           
           "Feedback_LabelsIncorrect", "Feedback_LabelsReversed",
           "Feedback_AllReal", "Feedback_AllFake",
           
           "AllRealConfidence","AllFakeConfidence" ,
           "Enjoyment")) {
  feedback_results <- rbind(feedback_results,
                            test_moderator(df, "Valence", m))
}

Warning in finalizeTMB(TMBStruc, obj, fit, h, data.tmb.old): Model convergence problem; non-positive-definite Hessian matrix. See vignette('troubleshooting')

Warning in finalizeTMB(TMBStruc, obj, fit, h, data.tmb.old): Model convergence problem; singular convergence (7). See vignette('troubleshooting'), help('diagnose')

display(filter(feedback_results, Outcome == "Valence"))

Parameter	Coefficient	95% CI	z	p	Moderator	Outcome
Gender [Male] × ConditionAI-Generated × Feedback NoFacesAttractiveTrue	0.06	[-0.17, 0.29]	0.51	0.611	Feedback_NoFacesAttractive	Valence
Gender [Female] × ConditionAI-Generated × Feedback NoFacesAttractiveTrue	-1.47e-03	[-0.19, 0.19]	-0.02	0.988	Feedback_NoFacesAttractive	Valence
Gender [Male] × ConditionAI-Generated × Feedback SomeFacesAttractiveTrue	0.03	[-0.06, 0.13]	0.64	0.521	Feedback_SomeFacesAttractive	Valence
Gender [Female] × ConditionAI-Generated × Feedback SomeFacesAttractiveTrue	0.06	[-0.05, 0.16]	1.10	0.270	Feedback_SomeFacesAttractive	Valence
Gender [Male] × ConditionAI-Generated × Feedback AIMoreAttractiveTrue	0.13	[ 0.01, 0.25]	2.13	0.033	Feedback_AIMoreAttractive	Valence
Gender [Female] × ConditionAI-Generated × Feedback AIMoreAttractiveTrue	0.06	[-0.05, 0.18]	1.05	0.293	Feedback_AIMoreAttractive	Valence
Gender [Male] × ConditionAI-Generated × Feedback AILessAttractiveTrue	-0.10	[-0.26, 0.06]	-1.19	0.234	Feedback_AILessAttractive	Valence
Gender [Female] × ConditionAI-Generated × Feedback AILessAttractiveTrue	0.05	[-0.12, 0.22]	0.55	0.579	Feedback_AILessAttractive	Valence
Gender [Male] × ConditionAI-Generated × Feedback AIMoreArousingTrue	-0.06	[-0.23, 0.11]	-0.68	0.497	Feedback_AIMoreArousing	Valence
Gender [Female] × ConditionAI-Generated × Feedback AIMoreArousingTrue	-0.13	[-0.42, 0.15]	-0.93	0.352	Feedback_AIMoreArousing	Valence
Gender [Male] × ConditionAI-Generated × Feedback AILessArousingTrue	0.02	[-0.17, 0.20]	0.19	0.848	Feedback_AILessArousing	Valence
Gender [Female] × ConditionAI-Generated × Feedback AILessArousingTrue	-0.04	[-0.23, 0.16]	-0.37	0.714	Feedback_AILessArousing	Valence
Gender [Male] × ConditionAI-Generated × Feedback DiffObviousTrue	-0.01	[-0.21, 0.18]	-0.13	0.898	Feedback_DiffObvious	Valence
Gender [Female] × ConditionAI-Generated × Feedback DiffObviousTrue	-0.11	[-0.36, 0.15]	-0.83	0.406	Feedback_DiffObvious	Valence
Gender [Male] × ConditionAI-Generated × Feedback DiffNoneTrue	6.34e-03	[-0.09, 0.10]	0.13	0.895	Feedback_DiffNone	Valence
Gender [Female] × ConditionAI-Generated × Feedback DiffNoneTrue	0.07	[-0.03, 0.17]	1.31	0.189	Feedback_DiffNone	Valence
Gender [Male] × ConditionAI-Generated × Feedback LabelsIncorrectTrue	-0.11	[-0.22, 0.00]	-1.93	0.054	Feedback_LabelsIncorrect	Valence
Gender [Female] × ConditionAI-Generated × Feedback LabelsIncorrectTrue	0.02	[-0.09, 0.14]	0.42	0.672	Feedback_LabelsIncorrect	Valence
Gender [Male] × ConditionAI-Generated × Feedback LabelsReversedTrue	0.09	[-0.14, 0.32]	0.78	0.435	Feedback_LabelsReversed	Valence
Gender [Female] × ConditionAI-Generated × Feedback LabelsReversedTrue	0.09	[-0.16, 0.35]	0.71	0.479	Feedback_LabelsReversed	Valence
Gender [Male] × ConditionAI-Generated × Feedback AllRealTrue	0.07	[-0.03, 0.17]	1.37	0.172	Feedback_AllReal	Valence
Gender [Female] × ConditionAI-Generated × Feedback AllRealTrue	0.10	[ 0.00, 0.20]	1.93	0.054	Feedback_AllReal	Valence
Gender [Male] × ConditionAI-Generated × Feedback AllFakeTrue	0.18	[-0.02, 0.38]	1.79	0.073	Feedback_AllFake	Valence
Gender [Female] × ConditionAI-Generated × Feedback AllFakeTrue	-0.21	[-0.44, 0.02]	-1.76	0.078	Feedback_AllFake	Valence
Gender [Male] × ConditionAI-Generated × AllRealConfidence	-0.03	[-0.09, 0.03]	-1.05	0.294	AllRealConfidence	Valence
Gender [Female] × ConditionAI-Generated × AllRealConfidence	0.03	[-0.04, 0.10]	0.84	0.400	AllRealConfidence	Valence
Gender [Male] × ConditionAI-Generated × AllFakeConfidence	0.02	[-0.27, 0.32]	0.16	0.874	AllFakeConfidence	Valence
Gender [Female] × ConditionAI-Generated × AllFakeConfidence	0.11	[-0.19, 0.42]	0.73	0.467	AllFakeConfidence	Valence
Gender [Male] × ConditionAI-Generated × Enjoyment	0.04	[-0.04, 0.12]	1.02	0.309	Enjoyment	Valence
Gender [Female] × ConditionAI-Generated × Enjoyment	9.40e-03	[-0.04, 0.06]	0.35	0.725	Enjoyment	Valence

Believed

m <- glmmTMB::glmmTMB(Valence ~ Gender /Condition + (Condition| Participant) + (1|Stimulus),
                      data=filter(df, ConditionBelief == "True"), 
                      family=glmmTMB::ordbeta(),
                      control = glmmTMB::glmmTMBControl(parallel = 8))

results_table(m)

Fixed Effects

Parameter	Coefficient	SE	95% CI	z	p
(Intercept)	0.32	0.07	(0.18, 0.46)	4.46	< .001
Gender (Female)	-0.09	0.08	(-0.25, 0.08)	-1.03	0.305
Gender (Male) × ConditionAI-Generated	-0.24	0.04	(-0.32, -0.15)	-5.59	< .001
Gender (Female) × ConditionAI-Generated	-0.17	0.05	(-0.26, -0.08)	-3.79	< .001

for(m in c("BAIT_Visual", "BAIT_Text", "BAIT_ExpertAI",
           "BAIT_Positive", "BAIT_Negative", "PornFrequency", "SexualActivity", 
           
           "Feedback_DiffObvious", "Feedback_DiffNone",
           "Feedback_AIMoreArousing", "Feedback_AILessArousing",
           "Feedback_AIMoreAttractive", "Feedback_AILessArousing",
           "Feedback_LabelsIncorrect", "Feedback_LabelsReversed")) {
  moderator_believed <- rbind(moderator_believed,
                             test_moderator(filter(df, ConditionBelief == "True"), "Valence", m))
}

(filter(moderator_believed, p < .1))

                                                              Parameter Coefficient         95% CI         z      p               Moderator  Outcome
85              Gender [Female] × ConditionAI-Generated × PornFrequency        0.06  [-0.01, 0.13]      1.65  0.099           PornFrequency  Arousal
86               Gender [Female] × SexualActivitywithin the last 3 days        0.50  [ 0.02, 0.98]      2.03  0.042          SexualActivity  Arousal
79  Gender [Male] × ConditionAI-Generated × Feedback AIMoreArousingTrue        0.25  [-0.04, 0.54]      1.72  0.086 Feedback_AIMoreArousing  Arousal
821              Gender [Female] × SexualActivitywithin the last 3 days        0.53  [ 0.13, 0.94]      2.56  0.010          SexualActivity Enticing
834             Gender [Female] × ConditionAI-Generated × BAIT Negative       -0.07 [-0.12, -0.01]     -2.33  0.020           BAIT_Negative  Valence
836              Gender [Female] × SexualActivitywithin the last 3 days        0.37  [ 0.05, 0.70]      2.23  0.026          SexualActivity  Valence
740 Gender [Male] × ConditionAI-Generated × Feedback AILessArousingTrue   -1.30e-04  [-0.33, 0.33] -7.79e-04 > .999 Feedback_AILessArousing  Valence
742 Gender [Male] × ConditionAI-Generated × Feedback AILessArousingTrue   -1.30e-04  [-0.33, 0.33] -7.79e-04 > .999 Feedback_AILessArousing  Valence

Emotionality (combined factor)

Here we combined the data for arousal, valence and enticing into an emotionality factor.

pca <- principal_components(df[c("Valence", "Arousal", "Enticing")], n = 2)
pca

# Loadings from Principal Component Analysis (no rotation)

Variable |   PC1 |   PC2 | Complexity
-------------------------------------
Valence  | -0.89 |  0.44 |       1.46
Arousal  | -0.91 | -0.37 |       1.33
Enticing | -0.95 | -0.06 |       1.01

The 2 principal components accounted for 94.67% of the total variance of the original data (PC1 = 83.46%, PC2 = 11.21%).

df$Emotionality <- as.numeric(normalize(-predict(pca)$PC1))

emotionality <- df |> 
  mutate(Emotionality_Extreme = ifelse(Emotionality %in% c(0, 1), "Extreme", "Not extreme")) |>
  ggplot(aes(x=Emotionality, fill = Gender)) +
  geom_histogram(aes(color=Emotionality_Extreme), position = "identity", bins=20) +
  facet_grid(~Gender, scales="free_y") +
  scale_color_manual(values=c("red", "white"), guide="none") +
  scale_fill_manual(values=c("Male"="#E91E63", "Female"="#FF9800"), guide="none") +
  scale_y_sqrt(expand = c(0, 0)) +
  labs(title = "Emotionality") +
  theme_abyss() +
  theme(plot.title = element_text(hjust = 0.5, face="bold"))

emotionality

emotionality_desc <- plot_descriptive(df, "Emotionality")
emotionality_desc

Whole Sample

This model looks at the effect of Gender, StimuliType and Condition on Emotionality, accounting for random variability due to participants and items (i.e., random intercepts).

Females, compared to males, report significantly lower emotionality ratings for fotographs. Among males, couple stimuli are rated lower than individual stimuli.Among females, couple stimuli are rated higher than individual stimuli. For males, AI-generated individual stimuli are rated lower than human-generated ones. Females have lower ratings for AI-generated couple stimuli compared to AI-generated stimuli of individuals.

m_emotionality <- glmmTMB::glmmTMB(Emotionality ~ Gender /StimuliType/ Condition + (Condition| Participant) + (1|Stimulus),
                      data=df, family=glmmTMB::ordbeta(),
                      control = glmmTMB::glmmTMBControl(parallel = 8))

results_table(m_emotionality)

Fixed Effects

Parameter	Coefficient	SE	95% CI	z	p
(Intercept)	0.48	0.11	(0.27, 0.69)	4.55	< .001
Gender (Female)	-1.11	0.15	(-1.41, -0.81)	-7.28	< .001
Gender (Male) × StimuliTypeCouple	-0.41	0.11	(-0.63, -0.19)	-3.68	< .001
Gender (Female) × StimuliTypeCouple	0.62	0.11	(0.40, 0.84)	5.44	< .001
Gender (Male) × StimuliTypeIndividual × ConditionAI-Generated	-0.11	0.04	(-0.19, -0.04)	-3.12	0.002
Gender (Female) × StimuliTypeIndividual × ConditionAI-Generated	0.04	0.04	(-0.04, 0.12)	0.94	0.348
Gender (Male) × StimuliTypeCouple × ConditionAI-Generated	-0.06	0.04	(-0.13, 6.10e-03)	-1.79	0.074
Gender (Female) × StimuliTypeCouple × ConditionAI-Generated	-0.06	0.04	(-0.14, 0.01)	-1.60	0.109

marginaleffects::avg_slopes(m_emotionality, variables="Condition", by="Gender", re.form=NA) |>
  results_table()

Comparison	Gender	Coefficient	z	p	95% CI
AI-Generated - Photograph	Male	-0.02	-3.13	0.002	[-0.04, -0.01]
AI-Generated - Photograph	Female	-3.34e-03	-0.45	0.650	[-0.02, 0.01]

Moderators

Looking at AI-attitudes, porn frequency and sexual activity, and feedback as moderators, we can see that for males reporting higher expertise in AI led to an increase of emotionality ratings for images labelled as AI-Generated. Additionally, males who reported having more negative attitiudes towards AI, rated AI-Generated images lower in emotionality compared to fotographs. While for feamles, those with more positive AI attitudes rate AI-generated images higher in emotionality.

Looking at feedback as a moderator, the results show that for males who reported finidng AI-generated images as more attractive rated these images higher in emotionality.Conversely, males who reported finding AI-generated images as less attractive rated these images lower in emotionality.

for(m in c("BAIT_Visual", "BAIT_Text", "BAIT_ExpertAI",
           "BAIT_Positive", "BAIT_Negative", "PornFrequency", "SexualActivity_num")) {
  moderator_results <- rbind(moderator_results,
                             test_moderator(df, "Emotionality", m))
}

display(filter(moderator_results, Outcome == "Emotionality"))

Parameter	Coefficient	95% CI	z	p	Moderator	Outcome
Gender [Male] × ConditionAI-Generated × BAIT Visual	0.05	[ 0.00, 0.11]	1.82	0.069	BAIT_Visual	Emotionality
Gender [Female] × ConditionAI-Generated × BAIT Visual	5.41e-03	[-0.05, 0.06]	0.18	0.854	BAIT_Visual	Emotionality
Gender [Male] × ConditionAI-Generated × BAIT Text	-3.70e-03	[-0.05, 0.04]	-0.16	0.872	BAIT_Text	Emotionality
Gender [Female] × ConditionAI-Generated × BAIT Text	0.01	[-0.04, 0.06]	0.53	0.595	BAIT_Text	Emotionality
Gender [Male] × ConditionAI-Generated × BAIT ExpertAI	0.04	[ 0.00, 0.07]	2.11	0.035	BAIT_ExpertAI	Emotionality
Gender [Female] × ConditionAI-Generated × BAIT ExpertAI	0.03	[-0.01, 0.07]	1.39	0.166	BAIT_ExpertAI	Emotionality
Gender [Male] × ConditionAI-Generated × BAIT Positive	0.03	[-0.02, 0.08]	1.08	0.282	BAIT_Positive	Emotionality
Gender [Female] × ConditionAI-Generated × BAIT Positive	0.05	[ 0.00, 0.09]	2.06	0.039	BAIT_Positive	Emotionality
Gender [Male] × ConditionAI-Generated × BAIT Negative	-0.04	[-0.07, -0.01]	-2.55	0.011	BAIT_Negative	Emotionality
Gender [Female] × ConditionAI-Generated × BAIT Negative	-0.02	[-0.06, 0.02]	-0.83	0.404	BAIT_Negative	Emotionality
Gender [Male] × ConditionAI-Generated × PornFrequency	5.80e-03	[-0.03, 0.04]	0.34	0.732	PornFrequency	Emotionality
Gender [Female] × ConditionAI-Generated × PornFrequency	-1.34e-03	[-0.04, 0.04]	-0.06	0.950	PornFrequency	Emotionality
Gender [Male] × ConditionAI-Generated × SexualActivity num	-0.03	[-0.06, 0.01]	-1.48	0.140	SexualActivity_num	Emotionality
Gender [Female] × ConditionAI-Generated × SexualActivity num	-0.03	[-0.07, 0.01]	-1.54	0.123	SexualActivity_num	Emotionality

for(m in c("Feedback_NoFacesAttractive", "Feedback_SomeFacesAttractive",
           "Feedback_AIMoreAttractive", "Feedback_AILessAttractive",
           "Feedback_AIMoreArousing", "Feedback_AILessArousing", 
            
           "Feedback_DiffObvious", "Feedback_DiffNone",
           
           "Feedback_LabelsIncorrect", "Feedback_LabelsReversed",
           "Feedback_AllReal", "Feedback_AllFake",
           
           "AllRealConfidence","AllFakeConfidence" ,
           "Enjoyment")) {
  feedback_results <- rbind(feedback_results,
                            test_moderator(df, "Emotionality", m))
}

Warning in finalizeTMB(TMBStruc, obj, fit, h, data.tmb.old): Model convergence problem; singular convergence (7). See vignette('troubleshooting'), help('diagnose')

display(filter(feedback_results, Outcome == "Emotionality"))

Parameter	Coefficient	95% CI	z	p	Moderator	Outcome
Gender [Male] × ConditionAI-Generated × Feedback NoFacesAttractiveTrue	-6.16e-03	[-0.27, 0.26]	-0.05	0.964	Feedback_NoFacesAttractive	Emotionality
Gender [Female] × ConditionAI-Generated × Feedback NoFacesAttractiveTrue	-0.09	[-0.32, 0.15]	-0.72	0.470	Feedback_NoFacesAttractive	Emotionality
Gender [Male] × ConditionAI-Generated × Feedback SomeFacesAttractiveTrue	0.02	[-0.09, 0.13]	0.38	0.707	Feedback_SomeFacesAttractive	Emotionality
Gender [Female] × ConditionAI-Generated × Feedback SomeFacesAttractiveTrue	0.05	[-0.07, 0.18]	0.85	0.393	Feedback_SomeFacesAttractive	Emotionality
Gender [Male] × ConditionAI-Generated × Feedback AIMoreAttractiveTrue	0.17	[ 0.03, 0.30]	2.43	0.015	Feedback_AIMoreAttractive	Emotionality
Gender [Female] × ConditionAI-Generated × Feedback AIMoreAttractiveTrue	0.09	[-0.05, 0.22]	1.23	0.220	Feedback_AIMoreAttractive	Emotionality
Gender [Male] × ConditionAI-Generated × Feedback AILessAttractiveTrue	-0.19	[-0.38, 0.00]	-2.01	0.045	Feedback_AILessAttractive	Emotionality
Gender [Female] × ConditionAI-Generated × Feedback AILessAttractiveTrue	0.03	[-0.17, 0.22]	0.26	0.796	Feedback_AILessAttractive	Emotionality
Gender [Male] × ConditionAI-Generated × Feedback AIMoreArousingTrue	-0.03	[-0.23, 0.18]	-0.25	0.803	Feedback_AIMoreArousing	Emotionality
Gender [Female] × ConditionAI-Generated × Feedback AIMoreArousingTrue	-0.19	[-0.52, 0.14]	-1.13	0.258	Feedback_AIMoreArousing	Emotionality
Gender [Male] × ConditionAI-Generated × Feedback AILessArousingTrue	0.03	[-0.19, 0.24]	0.22	0.823	Feedback_AILessArousing	Emotionality
Gender [Female] × ConditionAI-Generated × Feedback AILessArousingTrue	0.04	[-0.19, 0.27]	0.34	0.735	Feedback_AILessArousing	Emotionality
Gender [Male] × ConditionAI-Generated × Feedback DiffObviousTrue	-0.11	[-0.34, 0.13]	-0.88	0.380	Feedback_DiffObvious	Emotionality
Gender [Female] × ConditionAI-Generated × Feedback DiffObviousTrue	-0.07	[-0.35, 0.21]	-0.50	0.615	Feedback_DiffObvious	Emotionality
Gender [Male] × ConditionAI-Generated × Feedback DiffNoneTrue	0.04	[-0.07, 0.15]	0.66	0.509	Feedback_DiffNone	Emotionality
Gender [Female] × ConditionAI-Generated × Feedback DiffNoneTrue	0.03	[-0.10, 0.15]	0.42	0.677	Feedback_DiffNone	Emotionality
Gender [Male] × ConditionAI-Generated × Feedback LabelsIncorrectTrue	-0.09	[-0.22, 0.04]	-1.32	0.188	Feedback_LabelsIncorrect	Emotionality
Gender [Female] × ConditionAI-Generated × Feedback LabelsIncorrectTrue	-0.02	[-0.15, 0.12]	-0.25	0.804	Feedback_LabelsIncorrect	Emotionality
Gender [Male] × ConditionAI-Generated × Feedback LabelsReversedTrue	0.20	[-0.07, 0.47]	1.46	0.144	Feedback_LabelsReversed	Emotionality
Gender [Female] × ConditionAI-Generated × Feedback LabelsReversedTrue	-0.04	[-0.34, 0.26]	-0.27	0.789	Feedback_LabelsReversed	Emotionality
Gender [Male] × ConditionAI-Generated × Feedback AllRealTrue	0.02	[-0.09, 0.14]	0.38	0.701	Feedback_AllReal	Emotionality
Gender [Female] × ConditionAI-Generated × Feedback AllRealTrue	0.06	[-0.06, 0.18]	0.95	0.342	Feedback_AllReal	Emotionality
Gender [Male] × ConditionAI-Generated × Feedback AllFakeTrue	0.22	[-0.02, 0.45]	1.82	0.068	Feedback_AllFake	Emotionality
Gender [Female] × ConditionAI-Generated × Feedback AllFakeTrue	-0.03	[-0.31, 0.24]	-0.23	0.815	Feedback_AllFake	Emotionality
Gender [Male] × ConditionAI-Generated × AllRealConfidence	-0.02	[-0.08, 0.03]	-0.79	0.427	AllRealConfidence	Emotionality
Gender [Female] × ConditionAI-Generated × AllRealConfidence	6.01e-03	[-0.06, 0.08]	0.17	0.866	AllRealConfidence	Emotionality
Gender [Male] × ConditionAI-Generated × AllFakeConfidence	-2.85e-04	[-0.27, 0.27]	-2.08e-03	0.998	AllFakeConfidence	Emotionality
Gender [Female] × ConditionAI-Generated × AllFakeConfidence	0.09	[-0.25, 0.43]	0.54	0.592	AllFakeConfidence	Emotionality
Gender [Male] × ConditionAI-Generated × Enjoyment	0.06	[-0.04, 0.15]	1.20	0.229	Enjoyment	Emotionality
Gender [Female] × ConditionAI-Generated × Enjoyment	0.02	[-0.04, 0.08]	0.72	0.469	Enjoyment	Emotionality

Believed

m <- glmmTMB::glmmTMB(Emotionality ~ Gender / Condition + (Condition| Participant) + (1|Stimulus),
                      data=filter(df, ConditionBelief == "True"), 
                      family=glmmTMB::ordbeta(),
                      control = glmmTMB::glmmTMBControl(parallel = 8))

results_table(m)

Fixed Effects

Parameter	Coefficient	SE	95% CI	z	p
(Intercept)	0.13	0.10	(-0.06, 0.32)	1.37	0.171
Gender (Female)	-0.16	0.10	(-0.36, 0.05)	-1.51	0.132
Gender (Male) × ConditionAI-Generated	-0.33	0.05	(-0.42, -0.24)	-7.04	< .001
Gender (Female) × ConditionAI-Generated	-0.15	0.05	(-0.25, -0.05)	-2.91	0.004

for(m in c("BAIT_Visual", "BAIT_Text", "BAIT_ExpertAI",
           "BAIT_Positive", "BAIT_Negative", "PornFrequency", "SexualActivity", 
           
           "Feedback_DiffObvious", "Feedback_DiffNone",
           "Feedback_AIMoreArousing", "Feedback_AILessArousing",
           "Feedback_AIMoreAttractive", "Feedback_AILessArousing",
           "Feedback_LabelsIncorrect", "Feedback_LabelsReversed")) {
  moderator_believed <- rbind(moderator_believed,
                             test_moderator(filter(df, ConditionBelief == "True"), "Emotionality", m))
}

(filter(moderator_believed, p < .1))

                                                              Parameter Coefficient         95% CI         z      p               Moderator      Outcome
85              Gender [Female] × ConditionAI-Generated × PornFrequency        0.06  [-0.01, 0.13]      1.65  0.099           PornFrequency      Arousal
86               Gender [Female] × SexualActivitywithin the last 3 days        0.50  [ 0.02, 0.98]      2.03  0.042          SexualActivity      Arousal
79  Gender [Male] × ConditionAI-Generated × Feedback AIMoreArousingTrue        0.25  [-0.04, 0.54]      1.72  0.086 Feedback_AIMoreArousing      Arousal
821              Gender [Female] × SexualActivitywithin the last 3 days        0.53  [ 0.13, 0.94]      2.56  0.010          SexualActivity     Enticing
834             Gender [Female] × ConditionAI-Generated × BAIT Negative       -0.07 [-0.12, -0.01]     -2.33  0.020           BAIT_Negative      Valence
836              Gender [Female] × SexualActivitywithin the last 3 days        0.37  [ 0.05, 0.70]      2.23  0.026          SexualActivity      Valence
740 Gender [Male] × ConditionAI-Generated × Feedback AILessArousingTrue   -1.30e-04  [-0.33, 0.33] -7.79e-04 > .999 Feedback_AILessArousing      Valence
742 Gender [Male] × ConditionAI-Generated × Feedback AILessArousingTrue   -1.30e-04  [-0.33, 0.33] -7.79e-04 > .999 Feedback_AILessArousing      Valence
851              Gender [Female] × SexualActivitywithin the last 3 days        0.55  [ 0.13, 0.97]      2.56  0.011          SexualActivity Emotionality

Summary

• Gender is a major predictor of arousal, enticement and valence, with females generally reporting lower ratings than males overall.
• Women prefer couple images (i.e.,significantly higher arousal, enticement and valence ratings); men prefer individual images (i.e.,significantly lower arousal, enticement and valence ratings).
• AI-generation reduces arousal for men, especially for individual images. AI-generation has no significant effect on arousal for women, whether images are of individuals or couples.

Moderators

• For women, positive attitudes towards AI moderated arousal and enticement, resulting in more favorable evaluations of AI-generated images. Negative attitudes moderated valence ratings, leading to less favorable evaluations.
• Negative attitudes towards AI moderated valence ratings for both genders
• Reporting having higher knowledge about AI moderated valence ratings for men
• Sexual activity moderated enticement ratings for females. Specifically, females who reported having sexual activity within the 3 previous days tended to rate AI images higher in enticement.
• Men’s perceptions of the attractiveness of AI-generated images moderated both arousal and valence ratings. More attractive AI images were rated as more arousing and valent.
• Female’s feedback that no images were attractive moderated arousal ratings.

Figures

pred1 <- rbind(
  modelbased::estimate_means(m_arousal, by=c("Gender", "StimuliType", "Condition"), backend = "emmeans") |>
  as.data.frame() |> 
  mutate(Outcome="Arousal"),
  modelbased::estimate_means(m_enticement, by=c("Gender","StimuliType",  "Condition"), backend = "emmeans") |>
  as.data.frame() |> 
  mutate(Outcome="Enticement"),
  modelbased::estimate_means(m_valence, by=c("Gender", "StimuliType", "Condition"), backend = "emmeans") |>
  as.data.frame() |> 
  mutate(Outcome="Valence")
)

# stars1a <- data.frame(label=c("**","***", "**", "***", "***", "***"),
#                     Gender=fct_relevel(c("Male", "Female", "Male", "Female", "Male", "Female"), "Male"),
#                     Outcome=c("Arousal","Arousal", "Enticement", "Enticement", "Valence", "Valence"),
#                     Condition=c("Photograph", "Photograph", "Photograph", "Photograph", "Photograph", "Photograph"), 
#                     x=c(1.5, 1.5, 1.5, 1.5, 1.5, 1.5), 
#                     y = c(0.53, 0.40, 0.6, 0.51, 0.62, 0.58))

stars1b <- data.frame(label=c("*"),
                    Gender=fct_relevel(c("Male"), "Male"),
                    Outcome=c("Arousal"),
                    x=1,
                    y=0.60)

stars1b <- data.frame(label=c("*","**", "*","*"),
                    Gender=fct_relevel(c("Male", "Male", "Male","Male"), "Male"),
                    Outcome=c("Arousal", "Enticement", "Enticement","Valence"),
                    x=c(1, 1, 2, 1),
                    y=c(0.60, 0.65, 0.56, 0.68))


p1 <- pred1 |>
  mutate(Gender = fct_relevel(Gender, "Male", "Female")) |> 
  ggplot(aes(x=StimuliType, y=Proportion)) +
  geom_hline(yintercept=0, color="darkgrey") +
  geom_line(aes(group=Condition, color=Condition), position = position_dodge(width=0.2), alpha=0.8, linewidth=0.4, key_glyph = "path") +
  geom_pointrange(aes(ymin = CI_low, ymax=CI_high, color=Condition), position = position_dodge(width=0.2), size=0.3) +
  geom_text(data=stars1b, aes(x=x, y=y, label=label), hjust=0.5, color="#424242", size=6) +
  geom_rect(data=data.frame(StimuliType = "Individual", Proportion=0.1, Gender=as.factor("Male"), Outcome="Arousal"),
            aes(xmin=0.8, xmax=1.2, ymin=0.35, ymax=0.7), color="#8BC34A", alpha=0, ) +
  geom_rect(data=data.frame(StimuliType = "Individual",Proportion=0.1, Gender=as.factor("Female"), Outcome="Arousal"),
            aes(xmin=0.8, xmax=1.2, ymin=0.2, ymax=0.5), color="#8BC34A", alpha=0) +
  facet_grid(Outcome~Gender, switch="y", scales ="free") +
  scale_y_continuous(limits=c(0, 1), labels=scales::percent) +
  scale_color_manual(values=c("AI-Generated"="#2196F3", "Photograph"="#F44336")) +
  theme_minimal() +
  theme(axis.title.y = element_blank(),
        axis.title.x = element_blank(),
        axis.text.y = element_text(size = 8),
        strip.placement = "outside",
        strip.background.x = element_rect(fill=c("lightgrey"), color=NA),
        strip.text.x = element_text(size = 10),
        strip.text.y = element_text(size = 10),
        axis.text.x = element_text(angle=25, hjust=1, vjust=1.1, size=9, color="black"),
        legend.text = element_text(size = 10),
        legend.position = ) +
  labs(color="Images presented as:")


p1

m1 <- glmmTMB::glmmTMB(Arousal ~ Gender / Condition + (Condition | Participant) + (1|Stimulus),data = df, 
                      family=glmmTMB::ordbeta(),
                      control = glmmTMB::glmmTMBControl(parallel = 8))
m2 <- glmmTMB::glmmTMB(Arousal ~ Gender / Condition + (Condition | Participant) + (1|Stimulus),
                      data=filter(df, ConditionBelief == "True"), 
                      family=glmmTMB::ordbeta(),
                      control = glmmTMB::glmmTMBControl(parallel = 8))
m3 <- glmmTMB::glmmTMB(Arousal ~ Gender / Condition  + (Condition | Participant) + (1|Stimulus),
                      data=filter(df, ConditionBelief == "True", Feedback_NoFacesAttractive=="True"), 
                      family=glmmTMB::ordbeta(),
                      control = glmmTMB::glmmTMBControl(parallel = 8))

Warning in finalizeTMB(TMBStruc, obj, fit, h, data.tmb.old): Model convergence problem; non-positive-definite Hessian matrix. See vignette('troubleshooting')

Warning in finalizeTMB(TMBStruc, obj, fit, h, data.tmb.old): Model convergence problem; singular convergence (7). See vignette('troubleshooting'), help('diagnose')

m4 <- glmmTMB::glmmTMB(Arousal ~ Gender / Condition  + (Condition | Participant) + (1|Stimulus),
                      data=filter(df, ConditionBelief == "True", Feedback_AIMoreAttractive=="True"), 
                      family=glmmTMB::ordbeta(),
                      control = glmmTMB::glmmTMBControl(parallel = 8))

Warning in finalizeTMB(TMBStruc, obj, fit, h, data.tmb.old): Model convergence problem; singular convergence (7). See vignette('troubleshooting'), help('diagnose')

stars2 <- data.frame(label=c("**", "***", "*", "*", "*"),
                    Gender=fct_relevel(c("Male", "Male", "Male", "Female", "Female")),
                    Type=c("All trials", "Believed trials" ,"'AI was more attractive'", "Believed trials", "'No faces were attractive'"),
                    y=c(0.52, 0.56, 0.66, 0.47, 0.39))

pred2 <- rbind(
  estimate_relation(m1) |> 
    mutate(Type = "All trials"),
  estimate_relation(m2) |> 
    mutate(Type = "Believed trials"),
  estimate_relation(m3) |> 
    mutate(Type = "'No faces were attractive'"),
  estimate_relation(m4) |> 
    mutate(Type = "'AI was more attractive'"))

p2 <- pred2 |> 
  mutate(Type = fct_relevel(Type, "All trials", "Believed trials", "'No faces were attractive'", "'AI was more attractive'")) |> 
  ggplot(aes(x=Type, y=Predicted)) +
  geom_hline(yintercept=0, color="darkgrey") +
  geom_line(aes(group=Condition, color=Condition), position = position_dodge(width=0.2), alpha=0.8, linewidth=0.1) +
  geom_bar(stat="identity", aes(fill=Condition), position = position_dodge(width=0.2), width=0.2, alpha=0.6) +
  geom_pointrange(aes(ymin = CI_low, ymax=CI_high, color=Condition), position = position_dodge(width=0.2), size=0.2) +
  geom_rect(data=data.frame(Type="All trials", Predicted=0.1, Gender=as.factor("Male")),
            aes(xmin=0.7, xmax=1.3, ymin=0.35, ymax=0.55), color="#8BC34A", alpha=0) +
  geom_rect(data=data.frame(Type="All trials", Predicted=0.1, Gender=as.factor("Female")),
            aes(xmin=0.7, xmax=1.3, ymin=0.30, ymax=0.48), color="#8BC34A", alpha=0) +
  geom_text(data=stars2, aes(y=y, label=label), hjust=0.5, size=3) +
  facet_grid(~Gender) +
  scale_y_continuous(labels=scales::percent) +
  scale_color_manual(values=c("AI-Generated"="#2196F3", "Photograph"="#F44336")) +
  scale_fill_manual(values=c("AI-Generated"="#2196F3", "Photograph"="#F44336")) +
  labs(y="Arousal",
       tag = "B") +
  theme_minimal() +
  theme(axis.title.x = element_blank(),
        axis.text.x = element_text(angle=25, hjust=1, vjust=1.1, size=9),
        axis.text.y = element_text(size = 8),
        axis.title.y = element_text(size = 10),
        strip.placement = "outside",
        strip.background.x = element_rect(fill=c("lightgrey"), color=NA),
        strip.text.x = element_text(size = 10),
        legend.position = "none")

p2

dfsub$Gender <- as.factor(dfsub$Gender)
dfsub$Gender <- fct_relevel(dfsub$Gender, "Male", "Female")
dfsub$BAIT_Visual = (dfsub$BAIT_ImagesRealistic + (1 - dfsub$BAIT_ImagesIssues) + dfsub$BAIT_ImitatingReality + dfsub$BAIT_EnvironmentReal) / 4


m6 <- glm(Feedback_NoFacesAttractive ~ Gender / BAIT_Visual, 
    data=mutate(dfsub, Feedback_NoFacesAttractive = ifelse(Feedback_NoFacesAttractive =="True", 1, 0)), 
    family="binomial")

results_table(m6)

Parameter	Log-Odds	SE	95% CI	z	p
(Intercept)	-1.74	1.36	(-4.88, 0.62)	-1.28	0.202
Gender (Female)	-0.50	1.66	(-3.67, 3.06)	-0.30	0.763
Gender (Male) × BAIT Visual	-0.56	0.52	(-1.60, 0.50)	-1.06	0.289
Gender (Female) × BAIT Visual	0.04	0.33	(-0.59, 0.74)	0.11	0.913

Notes/ Questions

1. Models for arousal, enticement, and valence had singularity issues when also adding condition as random slope for stimuli
2. ‘Other’ was removed due to the very small percentage of individuals identifying as such.
3. ‘Porn’ factor was not computed, instead sexual activity and frequency of watching porn were treated seperate.
4. Missing final figures and interpretation of belief as influencing emotional ratings of images
5. The calculations for the BAIt/GAAIS dimensions were based on the validation of the BAIT scale

BAIT_Videos was calculated by averaging participants’ agreement with the statement that current AI algorithms can generate very realistic videos and the inverse of their agreement with the statement that videos generated by AI have obvious problems that make them easy to spot as fake. Higher scores reflect stronger beliefs in the visual realism and technical reliability of AI-generated videos.

BAIT_Visual was computed as the average of four items: belief that AI algorithms can generate very realistic images, the inverse of belief that AI-generated images always contain errors and artifacts, agreement that CGI can perfectly imitate reality, and belief that technology allows the creation of environments that seem just as real as reality. Higher scores capture stronger beliefs in the visual realism and immersive quality of AI-generated and computer-generated content.

BAIT_Text was calculated by averaging agreement with the belief that AI assistants can write texts indistinguishable from those written by humans and the inverse of the belief that AI-written documents usually read differently than human-written ones. Higher scores indicate stronger beliefs in the linguistic realism of AI-generated text.

BAIT_Negative was computed by averaging levels of agreement with the statements “AI is dangerous” and “I am worried about future uses of AI”.This composite reflects participants’ negative attitudes toward AI.

BAIT_Positive was calculated by averaging agreement with the statements “AI is exciting” and “Much of society will benefit from a future full of AI”. This score represents positive attitudes towards AI.