# Load required libraries
library(dplyr)
library(tidyr)
library(purrr)
library(readr)
library(stringr)
library(jsonlite)
library(magrittr)
library(ggplot2)
library(effectsize)
# Step 1: Import and Combine Data
# Read all CSV files from the specified folder, extract subject IDs from file names,
# and combine them into one data frame.
folder_path <- "/Users/miladutton/Desktop/CSS/CSS204/critcher2013_2/data/participant_data"
csv_files <- list.files(path = folder_path, pattern = "*.csv", full.names = TRUE, recursive = TRUE)
combined_data <- csv_files %>%
map_dfr(~ {
data <- read_csv(.x)
subject_id <- tools::file_path_sans_ext(basename(.x)) # Extract subject ID from file name
data %>% mutate(subject_id = subject_id, .before = 1) # Add subject ID as the first column
})
# Step 2: Filter Relevant Columns and Add Condition
# 1. Group data by participant to calculate the maximum total time elapsed for each participant
# 2. Add a column (`total_time_elapsed`) to store the total time elapsed per participant.
# 3. Filter out participants whose total time elapsed is less than 1.5 minutes (90,000 ms).
# 4. Select only relevant columns for analysis.
# 5. Classify stimuli into conditions (moral or immoral) based on keywords in the stimulus text.
# 6. Fill down condition labels for each participant to propagate the classification.
cleaned_data <- combined_data %>%
group_by(subject_id) %>%
mutate(total_time_elapsed = max(time_elapsed, na.rm = TRUE)) %>% # Step 1 and 2
ungroup() %>%
filter(total_time_elapsed > (1.5 * 60 * 1000)) %>% # Step 3
select(subject_id, rt, stimulus, response, time_elapsed) %>% # Step 4
mutate(condition = case_when( # Step 5
str_detect(stimulus, "steal the money") ~ "moral",
str_detect(stimulus, "pocketed the money") ~ "immoral",
TRUE ~ NA_character_
)) %>%
fill(condition, .direction = "down")
# Step 3: Extract Attention Check
# Identify and extract attention check responses for each participant,
# filling them across rows within each subject group.
cleaned_data <- cleaned_data %>%
group_by(subject_id) %>%
mutate(attention_check = if_else(
row_number() == n() & str_detect(response, '"Q0":'),
str_extract(response, '(?<=\\\"Q0\\\":)\\d+'), # Extract Q0 value
NA_character_
)) %>%
fill(attention_check, .direction = "downup") %>%
ungroup() %>%
mutate(attention_check = as.numeric(attention_check)) # Convert to numeric
# Step 4: Parse JSON Responses
# Parse JSON responses, expand them into separate columns, and remove rows
# containing only attention check responses or empty question values.
split_data <- cleaned_data %>%
mutate(response_parsed = map(response, ~ {
tryCatch(fromJSON(.) %>% as.data.frame(), error = function(e) data.frame())
})) %>%
unnest(cols = response_parsed, keep_empty = TRUE) %>%
filter(
!str_detect(response, '^\\{"Q0":\\d+\\}$') & # Exclude attention-check-only rows
!if_all(starts_with("Q"), is.na) # Exclude rows with all Q columns as NA
)
# Step 5: Assign Targets (Justin/Nate)
# Alternate rows between "Justin" and "Nate" for target assignment.
split_data <- split_data %>%
mutate(target = if_else(row_number() %% 2 == 1, "Justin", "Nate"))
# Step 6: Filter for Valid Participants and Reshape Data to Long Format
# 1. Filter participants who pass attention checks and meet response criteria.
# 2. Reshape data into long format, keeping all questions for valid participants.
# 3. Exclude participants who gave the same response to every question.
# Reshape data to long format
long_data <- split_data %>%
pivot_longer(
cols = starts_with("Q"),
names_to = "question",
values_to = "response_value"
) %>%
mutate(
question_value = paste(target, question, sep = "_")
)
# Identify valid participants
valid_participants <- long_data %>%
filter(attention_check == 0) %>% # Ensure attention_check is always 0
filter(
(question_value == "Justin_Q0" & response_value == 1) | # Ensure Justin_Q0 is 1
(question_value == "Nate_Q0" & response_value == 0) # Ensure Nate_Q0 is 0
) %>%
group_by(subject_id) %>%
filter(n_distinct(response_value) > 1) %>% # Exclude participants with identical responses
ungroup() %>%
distinct(subject_id) %>%
pull(subject_id)
# Filter long_data for valid participants and enforce conditions
long_data <- long_data %>%
filter(subject_id %in% valid_participants) %>%
filter(
!(question_value == "Justin_Q0" & response_value != 1) & # Enforce Justin_Q0 = 1
!(question_value == "Nate_Q0" & response_value != 0) # Enforce Nate_Q0 = 0
)
# Step 7: Create Filtered Dataset with Specified Columns
# Keep only `subject_id`, `condition`, `response_value`, `question_value`, and `decision_speed`.
filtered_long_data <- long_data %>%
mutate(decision_speed = case_when(
str_detect(question_value, "Justin") ~ "quick",
str_detect(question_value, "Nate") ~ "slow",
TRUE ~ NA_character_
)) %>%
select(subject_id, condition, response_value, question_value, decision_speed)
# Step 8: Filter for Relevant Questions
# Keep only specific questions of interest and convert response values to numeric.
relevant_questions <- c("Justin_Q1", "Justin_Q2", "Justin_Q3", "Nate_Q1", "Nate_Q2", "Nate_Q3")
cleaned_long_data <- filtered_long_data %>%
filter(question_value %in% relevant_questions) %>%
mutate(response_value = as.numeric(response_value))
# Step 9: Calculate Composite Scores
# Compute average moral scores for Justin and Nate across their respective questions.
cleaned_long_data <- cleaned_long_data %>%
group_by(subject_id) %>%
mutate(
justin_moral_score = mean(response_value[question_value %in% c("Justin_Q1", "Justin_Q2", "Justin_Q3")], na.rm = TRUE),
nate_moral_score = mean(response_value[question_value %in% c("Nate_Q1", "Nate_Q2", "Nate_Q3")], na.rm = TRUE)
) %>%
ungroup()
# Step 10: Final Dataset
# Reshape data for analysis and include relevant attributes.
final_dataset <- cleaned_long_data %>%
select(subject_id, condition, justin_moral_score, nate_moral_score) %>%
distinct() %>%
pivot_longer(
cols = c(justin_moral_score, nate_moral_score),
names_to = "character",
values_to = "moral_score"
) %>%
mutate(
character = ifelse(character == "justin_moral_score", "Justin", "Nate"),
decision_speed = ifelse(character == "Justin", "quick", "slow")
) %>%
select(subject_id, condition, character, decision_speed, moral_score)
write.csv(final_dataset, "final_individual.csv", row.names = FALSE)Replication of Study ‘How Quick Decisions Illuminate Moral Character’ by Critcher et al. (2013, Social Psychological and Personality Science)
Introduction
This paper examines how the speed in making moral decisions can shape how others view as, as well as how we assess our own actions. The intersection between decision making and moral cognition touches on the core psychological questions regarding ethics, thought process, and social environments.
The experiment will follow the structure of the original study by having participants read stories regarding making moral decisions at varying speeds. In experiment 1, two men find separate wallets in a parking lot. One man immediately decides what to do with the wallet while the other decides after a “long and careful deliberation”. In the moral condition, both men turn the wallet in, and in the immoral decision both men keep the wallet. Participants are then given questionnaires in which they rank the following sets: quickness, moral character evaluation, certainty, and emotional impulsivity. In this study, I only replicated experiment 1. For analysis, A two way ANOVA test was run to assess decision speed and type of decision on moral character evaluation.
Methods
Power Analysis
The original study utilizes a series of analyses to determine the interaction between dependent and independent variables.
Decision on Moral Character: A significant main effect of decision type (F(1, 117) = 541.52, p < .001) revealed that Justin, who decided quickly (M = 6.44, SE = .08), was perceived differently from Nate (M = 2.15, SE = .12). Moral decision also influenced character evaluation (F(1, 117) = 127.07, p < .001), with returning the wallet seen as morally better.
Decision Speed and Emotional Impulsivity: Justin was viewed as less emotionally impulsive (M = 2.40, SE = .11) than Nate (M = 3.79, SE = .12), (F(1, 117) = 95.26, p < .001), but this did not affect moral evaluations (t < 1).
Decision Speed and Moral Evaluation Polarization: The Decision × Speed interaction was significant (F(1, 117) = 127.07, p < .001), with quick immoral decisions leading to more negative evaluations (t(54) = 8.28, p < .001) and quick moral decisions to more positive evaluations (t(63) = 7.71, p < .001).
Certainty as a Mediator: Quick decisions were associated with higher certainty (F(1, 117) = 706.6, p < .001). Certainty mediated the relationship between decision speed and moral evaluation, with greater certainty explaining more negative evaluations for immoral decisions and more positive evaluations for moral ones.
I planned to perform power analyses to determine the sample sizes needed to achieve 80%, 90%, 95% power for detecting effect sizes from the original study. However, the original study did not give exact effect sizes, difference between means, or error bars. For that reason, I was unable to perform a power analysis and determine sample size through that methodology.
Planned Sample
Based on the original studies findings, I planned to have a sample size of approximately 298 participants. This is 2.5 times the original sample size n = 119. However, due to funding I ended up doing 1/3 of that, with a sample size of n = 100, and the combining data with the other two groups who replicated the same study. The sampling frame consisted of individuals in the participant pool on the website Prolific. The only pre selection rule was that participants must be fluent in English and be approved for at least 98% of the studies they applied to.
Materials
Participants in this study were presented with a story focusing on two men, as outlined in the original article. Through reaching out to the original researchers, I was able to get access to their exact materials. The story was structured as follows.
“Justin and Nate were walking to separate cars in the parking lot of Andronico’s, a local grocery store, when they each spotted a different lost wallet next to their cars. Upon picking up the wallet and looking inside, each found several hundred dollars in cash. Each man considered whether he should return the wallet and money to the Customer Service desk at Andronico’s, or pocket the money and drive away.”
Moral Condition
“Justin saw his decision as an easy one and was able to decide quickly. He did not steal the money and instead left the wallet with Customer Service.”
“In contrast, Nate saw his decision as difficult and was only able to decide after long and careful deliberation. After several minutes of thinking in his car, he too decided not to steal the money and instead left the wallet with Customer Service.”
Immoral Condition
“Justin saw his decision as an easy one and was able to decide quickly. He pocketed the money and drove off.”
“In contrast, Nate saw his decision as difficult and was only able to decide after long and careful deliberation. After several minutes of thinking in his car, he too decided to pocket the money and drive off.”
Immediately following the description of Justin and Nate’s actions, participants answered a series of questions related to the men’s moral character evaluation. These questions were rated on a 1-7 scale. These questions were also taken directly from the original research.
Quickness (manipulation check):
- Q0: Did [Character] make his decision quickly or slowly? (1 = particularly slowly, 7 = particularly quickly)
Moral Character Evaluation:
- Q1: Regardless of [Character]’s decision, does it sound like [Character] has underlying moral principles that are good, bad, or somewhere in between? (1 = completely bad, 4 = mixed, 7 = completely good)
- Q2: Regardless of [Character]’s decision, do you think Justin has moral standards that are good, bad, or somewhere in between? (1 = completely bad, 4 = mixed, 7 = completely good)
- Q3: Regardless of [Character]‘s decision, do you think Justin possesses the moral knowledge and principles necessary to do ‘the right thing’? (1 = not at all, 4 = somewhat, 7 = completely)
Certainty:
- Q4: Would you say [Character] was quite certain in his decision, or did [Character] have hesitations about his or her decision? (1 = completely certain, 7 = considerable hesitations)
- Q5: How close do you think [Character] was to choosing the alternate course of action? (1 = very close to, 7 = not close at all)
- Q6: How conflicted do you think [Character] felt in making the decision? (1 = very conflicted, 7 = not at all conflicted)
- Q7: Based on the information provided, do you think [Character] had many reservations about the decision? (1 = none at all, 7 = a whole lot)
Emotional impulsivity:
- Q8: Do you think [Character] was calm and emotionally contained while making the decision? (1 = not at all, 7 = entirely so)
- Q9: To what extent do you think [Character] became upset and acted without thinking? (1 = not at all, 7 = entirely so)
Note:
- These materials were used precisely as outlined in the original article without deviations.
Procedure
Participants are randomly assigned to the moral or immoral decision.
Moral decision: Justin and Nate both decide not to steal the money and return the wallets.
Immoral decision: Justin and Nate both decide to keep the money and drive off.
Participants read about both Justin and Nate, who found wallets in the parking lot. “Justin ‘was able to decide quickly’ what to do. Nate ‘was only able to decide after long and careful deliberation.’”
Participants rate their quickness, moral character evaluation, certainty, and emotional impulsivity on a likert scale.
Participants answer attention check to confirm their attention in the study.
Analysis Plan
The main focus of this study is to evaluate moral character judgments based on decision type and decision speed. After collecting the data, I excluded participants who met the exclusion criteria.
For the statistical analysis, my primary question of interest involved testing the effect of two factors (decision speed and decision type) on moral character judgments. A two-way ANOVA was conducted to evaluate the main effects and interaction effects. Effect sizes, such as partial eta-squared, were calculated to interpret the magnitude of these effects.
The results of the ANOVA were visualized using several approaches. Bar plots with error bars were used to compare mean moral scores across conditions, providing a clear depiction of differences between groups. Box plots highlighted the distribution of moral scores within each condition, revealing variability and potential outliers. Interaction plots illustrated how the relationship between decision speed and moral character evaluation varied across decision types, offering a visual representation of significant interaction effects. These methods allowed for a comprehensive and nuanced interpretation of the data.
Design Overview
Two factors were manipulated: decision type and decision speed.
Four measures were taken: quickness, moral character evaluation, certainty, emotional impulsivity.
It was a within subjects design.
Measures were not repeated.
There was no mention of steps taken to reduce demand characteristics.
Participants with previous traumatic exposure to a relevant situation or personal ties may have strong feelings regarding moral evaluation, which may serve as a confounding variable.
Differences from Original Study
The main difference from the original study is that I performed one experiment rather than two. Specifically, I replicated experiment 1.
The participant pool was also different. The original paper included mainly undergraduate students at UC Berkeley, as well as members from the surrounding community. As I am not a part of that institution, my participants were not from that community. I conducted this study entirely online, unlike the original study, which did not specify whether participants were tested in a lab or online. Using Prolific, I did not impose any regional restrictions, which meant I could not control the exact environment in which participants completed the study.
Lastly, I added an attention check at the end of the questions which the original study did not have. This was to ensure that participants were not just clicking through, and to help minimize bots. These differences were not expected to make a difference in obtaining the effect.
Methods Addendum (Post Data Collection)
After collecting and cleaning my data, I decided to handle missing values through list wise deletion. This approach, which removes participants with missing data for any variable included in the analysis, was not part of my original plan but was implemented to maintain a consistent data set for all statistical tests. While this method reduced the sample size slightly, it ensured that all analyses were conducted on complete cases, preserving the validity of the results.
Actual Sample
After cleaning the data and excluding invalid participants, I had a sample size of n = 96. After combing my data with the other groups and cleaning it, I had a final sample size of n = 207. I did not collect any demographic information.
Differences from pre-data collection methods plan
None.
Results
Data preparation
To prepare the data set for analysis, I implemented a detailed cleaning process to ensure its integrity and reliability. A manual check for missing values were removed to maintain data quality, however none were found. Participants who met any one part the exclusion criteria were also removed.
The exclusion criteria is as follows:
Failed the manipulation check
Failed the attention check
Completed the study in less than 1.5 minutes
Had the same response to every question
A condition column was added to categorize scenarios into moral or immoral decisions. Responses were parsed to extract structured data from JSON format, handling invalid responses to avoid errors. The data set was then reshaped into a long format, which allowed for easier analysis. Composite scores for moral evaluations were calculated for each character, providing a clear and structured summary of the data. These steps ensured the data set was properly formatted, high-quality, and ready for statistical analysis and visualization.
My data set was then combined with the data from the other two groups. Due to different cleaning techniques some of the original structure from my data set was lost in integration with the new one. This new combined data set had missing values, so those were dropped. The mean moral composite score was recalculated again and columns for speed and character were added.
# Import and clean the combined groups dataset
# Define the folder path
folder_path <- "/Users/miladutton/Desktop/CSS/CSS204/critcher2013_2/data/Datasets"
# Read the dataset
quick_decisions_combined <- read.csv(file.path(folder_path, "quick_decisions_combined.csv"))
# Drop rows with missing values for any column
cleaned_combined <- quick_decisions_combined %>%
drop_na()
# Calculate mean moral scores for Justin (questions 2-4) and Nate (questions 2-4)
# Note shift in question numbering from my original data - questions now start at 1 rather than 0
mean_combined <- cleaned_combined %>%
mutate(
Justin_mean_moral = rowMeans(select(., Justin_Q2:Justin_Q4), na.rm = TRUE),
Nate_mean_moral = rowMeans(select(., Nate_Q2:Nate_Q4), na.rm = TRUE)
)
# Reshape the data to long format
final_combined <- mean_combined %>%
select(condition, Justin_mean_moral, Nate_mean_moral) %>%
pivot_longer(
cols = c(Justin_mean_moral, Nate_mean_moral),
names_to = "character",
values_to = "moral_score"
) %>%
mutate(
decision_speed = if_else(character == "Justin_mean_moral", "quick", "slow"),
character = if_else(character == "Justin_mean_moral", "Justin", "Nate")
) %>%
arrange(character, condition)
# Save the final dataset
write.csv(final_combined, "final_combined.csv", row.names = FALSE)Confirmatory analysis
A two-way ANOVA was conducted to examine the main effects of decision type (whether participants decided to return or keep the wallet) and decision speed (quick or slow) on the moral character evaluation. This analysis allowed me to assess how these factors independently influenced moral judgments, both independently and when interacting.
# ANOVA for just my group data
# Perform Two-Way ANOVA
anova_results <- aov(moral_score ~ condition * decision_speed, data = final_dataset)
# Display the ANOVA table
summary(anova_results) Df Sum Sq Mean Sq F value Pr(>F)
condition 1 180.66 180.66 188.942 <2e-16 ***
decision_speed 1 0.06 0.06 0.061 0.806
condition:decision_speed 1 114.50 114.50 119.743 <2e-16 ***
Residuals 188 179.76 0.96
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Above are the ANOVA results for my individual sample, n=96.
Condition (F = 188.94, p < 0.001): A significant main effect of condition shows that moral evaluation scores vary substantially between moral and immoral conditions.
Decision Speed (F = 0.06, p = 0.806): Decision speed alone does not have a significant effect on moral evaluation scores.
Condition × Decision Speed Interaction (F = 119.74, p < 0.001): A significant interaction effect indicates that decision speed moderates the relationship between condition and moral evaluation
# Bar Plot for my groups data
plot <- ggplot(final_dataset, aes(x = condition, y = moral_score, fill = decision_speed)) +
stat_summary(
fun = "mean",
geom = "bar",
position = position_dodge(0.7),
color = "black",
width = 0.6
) +
stat_summary(
fun.data = "mean_cl_boot",
geom = "linerange",
position = position_dodge(0.7),
linewidth = 0.8
) +
scale_fill_manual(values = c("gray", "blue")) +
labs(
title = "Moral Character Evaluation - Individual Sample",
x = "Condition",
y = "Mean Moral Score",
fill = "Decision Speed"
) +
theme_minimal() +
theme(
panel.background = element_rect(fill = "white", color = NA), # Ensure white background
plot.background = element_rect(fill = "white", color = NA) # Ensure white plot background
)
# Display
print(plot)
Above are the plotted results for my individual sample, n=96.
# ANOVA for just combined groups data
# Run the two-way ANOVA
anova_result_combined <- aov(moral_score ~ condition * decision_speed, data = final_combined)
# Display the summary of the ANOVA
summary(anova_result_combined) Df Sum Sq Mean Sq F value Pr(>F)
condition 1 325.8 325.8 315.4 <2e-16 ***
decision_speed 1 0.0 0.0 0.0 0.987
condition:decision_speed 1 198.1 198.1 191.8 <2e-16 ***
Residuals 410 423.5 1.0
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Above are the ANOVA results for the combined sample, n = 207.
Condition (F = 315.4, p < 0.001): A strong and significant main effect of condition indicates that the moral evaluation scores differ significantly between the moral and immoral conditions.
Decision Speed (F = 0.0, p = 0.987): Decision speed alone does not significantly affect moral evaluation scores.
Condition × Decision Speed Interaction (F = 191.8, p < 0.001): A significant interaction effect indicates that the influence of condition on moral evaluation depends on the level of decision speed.
#Bar plot for combined groups data
library(ggplot2)
# Bar Plot for the groups data
plot <- ggplot(final_combined, aes(x = condition, y = moral_score, fill = decision_speed)) +
stat_summary(
fun = "mean",
geom = "bar",
position = position_dodge(0.7),
color = "black",
width = 0.6
) +
stat_summary(
fun.data = "mean_cl_boot",
geom = "linerange",
position = position_dodge(0.7),
linewidth = 0.8
) +
scale_fill_manual(values = c("gray", "blue")) +
labs(
title = "Moral Character Evaluation - Combined Sample",
x = "Condition",
y = "Mean Moral Score",
fill = "Decision Speed"
) +
theme_minimal() +
theme(
panel.background = element_rect(fill = "white", color = NA),
plot.background = element_rect(fill = "white", color = NA)
)
# Display the plot
print(plot)
Above are the plotted results for the combined sample, n=207.
Comparison of replicated and original graph.
Exploratory analyses
#Exploring the effect size for my groups data
effect_sizes <- eta_squared(anova_results, partial = TRUE)
print(effect_sizes)# Effect Size for ANOVA (Type I)
Parameter | Eta2 (partial) | 95% CI
--------------------------------------------------------
condition | 0.50 | [0.42, 1.00]
decision_speed | 3.22e-04 | [0.00, 1.00]
condition:decision_speed | 0.39 | [0.30, 1.00]
- One-sided CIs: upper bound fixed at [1.00].Above are the effect size results for my individual sample, n=96.
Condition (η² = 0.50): There is a large effect size for the condition variable, suggesting that the condition (e.g., moral vs. immoral) has a significant impact on the moral evaluation scores.
Decision Speed (η² ≈ 0.00): The effect size for decision speed is negligible, indicating that whether decisions were made quickly or slowly does not significantly influence moral evaluations in this sample.
Condition × Decision Speed Interaction (η² = 0.39): A moderate to large interaction effect suggests that the relationship between condition and moral evaluation is influenced by decision speed to a large extent.
#Exploring the effect size for combined data
effect_sizes_combined <- eta_squared(anova_result_combined, partial = TRUE)
print(effect_sizes_combined)# Effect Size for ANOVA (Type I)
Parameter | Eta2 (partial) | 95% CI
--------------------------------------------------------
condition | 0.43 | [0.38, 1.00]
decision_speed | 6.34e-07 | [0.00, 1.00]
condition:decision_speed | 0.32 | [0.26, 1.00]
- One-sided CIs: upper bound fixed at [1.00].Above are the effect size results for the combined sample, n=207.
Condition (η² = 0.43): The condition variable remains a strong predictor of moral evaluation, though the effect size is slightly lower compared to the individual sample.
Decision Speed (η² ≈ 0.00): Similar to the individual sample, decision speed has an almost negligible effect on moral evaluations.
Condition × Decision Speed Interaction (η² = 0.32): The interaction effect is slightly smaller than in the individual sample but still demonstrates a moderate influence.
# Interaction plot
interaction_plot <- ggplot(final_combined, aes(x = condition, y = moral_score, color = decision_speed, group = decision_speed)) +
geom_point(size = 4, shape = 21, fill = "white") +
geom_line(linewidth = 1.2) +
scale_color_manual(values = c("darkgrey", "blue")) +
labs(
title = "Effect of Condition on Moral Evaluation by Decision Speed",
x = "Condition",
y = "Moral Score",
color = "Decision Speed"
) +
theme_minimal(base_size = 14) +
theme(
plot.title = element_text(hjust = 0.5, face = "bold", size = 16),
axis.title.x = element_text(size = 14, face = "italic"),
axis.title.y = element_text(size = 14, face = "italic"),
axis.text = element_text(size = 12),
legend.position = "top",
legend.title = element_text(face = "bold"),
panel.grid.major = element_line(color = "gray85", size = 0.5) # Subtle gridlines
)
interaction_plot
This interaction plot (n=207) demonstrates a notable interaction between condition (moral vs. immoral) and decision speed (quick vs. slow). Specifically:
Quick Decisions: There is a sharper increase in moral scores for the moral condition compared to the immoral condition.
Slow Decisions: The change in moral scores is less pronounced, indicating that decision speed modulates the influence of condition on moral evaluations.
The overall pattern highlights that decision speed impacts the degree to which moral and immoral conditions are evaluated differently.
ggplot(final_combined, aes(x = condition, y = moral_score, fill = decision_speed)) +
geom_boxplot() +
scale_fill_manual(values = c("grey", "blue")) +
labs(
title = "Moral Scores by Condition and Character",
x = "Condition",
y = "Moral Score",
fill = "Speed"
) +
theme_minimal(base_size = 14) +
theme(
plot.title = element_text(hjust = 0.5, face = "bold", size = 16),
axis.title.x = element_text(size = 14, face = "italic"),
axis.title.y = element_text(size = 14, face = "italic"),
axis.text = element_text(size = 12),
legend.position = "top"
)
This box plot (n=207) further illustrates the interaction:
Immoral Condition: Quick decision speeds result in lower variability and slightly lower moral scores compared to slow decision speeds.
Moral Condition: The slow decision speed group demonstrates greater variability in moral scores, but the mean scores remain higher compared to the immoral condition.
Discussion
Summary of Replication Attempt
The primary result of the confirmatory analysis in this replication showed a significant interaction effect between decision type (whether participants decided to return or keep the wallet) and decision speed (quick vs. slow) on moral character evaluation, consistent with the original findings. Specifically, the interaction effect was significant (F(1, 410) = 191.8, p < 2e-16), confirming that quick moral decisions received more positive evaluations than slow moral decisions, and quick immoral decisions were judged more harshly than slow immoral decisions. This finding is comparable to the original finding of F(1, 117) = 127.07, p < .001.
Similarly, the original paper also found a significant effect of decision type on moral character evaluation. In the original study, the main effect of decision type was significant (F(1, 117) = 417.07, p < .001) and explained much of the variance in moral character evaluations, demonstrating that moral decisions were evaluated more positively than immoral ones. In the replication, the effect of decision type was also significant (F(1, 410) = 315.4, p < 2e-16) and similarly strong, confirming that moral decisions received higher moral character evaluations compared to immoral ones.
This replication successfully captured the ideas of the original finding, which demonstrated that decision speed influenced the perceived moral character of individuals, with quick decisions being interpreted as reflecting greater certainty and thus more polarized moral evaluations. Therefore, the replication can be categorized as a partial replication of the original result, as the primary interaction effect was consistent, but additional analyses, such as emotional impulsivity or mediation through perceived certainty, were not evaluated.
Commentary
Insights from Follow-Up Exploratory Analysis
In the exploratory analyses, the replication confirmed the interaction effect between decision type and decision speed, consistent with the original study. However, the non significant effect of decision speed (F(1, 410) = 0.0, p = 0.987) suggests that the speed of the decision alone does not influence moral character evaluation. Decision speed only has a significant effect on moral character evaluation when interacting with decision type. This highlights that the moral nature of the decision is a stronger determinant of judgments than the speed at which the decision is made.
Assessment of the Meaning of the Replication
This replication is successful in reproducing the key findings of the original study, particularly the significant interaction between decision type and speed. However, slight differences in effect sizes could be attributed to the larger sample size in this study (n = 207 compared to n = 118 in the original). While the replication used the same moral character evaluation measures, minor procedural differences, such as the participant pool, experimental environment, or even cultural contexts, could explain any small differences. Importantly, the replicated findings strengthen the argument that quick decisions are perceived as more certain and thus more polarizing in moral contexts.
Objections or Challenges Raised:
One potential challenge in the replication is the decision to handle missing data through list wise deletion, which could reduce variability in the data set. Additionally, while the replication focused solely on decision type and speed, the original study may have included mediating variables (e.g., perceived certainty or emotional impulsivity) that were not tested here. No direct objections from the original authors have been raised.
CREDIT and acknowledgements
Mila Dutton: Software (lead); formal analysis (lead); writing - original draft (equal); writing – review and editing (equal). Caroline Porche: Methodology (lead); software (support); writing - original draft (equal); writing – review and editing (equal). Anthony Zavala: Methodology (support); Writing - original draft (equal); writing – review and editing (equal).