Survey results¶
Introduction¶
Here we will analyze and visualize the survey data.
Initial setup¶
Install and load the needed libraries and the survey data.
In [ ]:
# List of packages to install
packages <- c("tidyverse", "syuzhet", "wordcloud", "tidytext", "readxl", "tm")
# Function to check and install packages if not already installed
install_if_missing <- function(p) {
if (!requireNamespace(p, quietly = TRUE)) {
install.packages(p)
}
}
# Install necessary packages
invisible(sapply(packages, install_if_missing))
# Load necessary libraries
library(tidyverse) # for data manipulation
library(syuzhet) # for sentiment analysis
library(wordcloud) # for word clouds
library(tidytext) # for text mining
library(readxl) # for reading excel files
library(tm) # for text mining
# Load the provided survey data
file_path <- "data/cleanSurveyData20240514.xlsx"
# this line can be repeated later to reset the data
data <- read_excel(file_path)
# Suppress warnings (this is fine)
options(warn = -1)
Show the column names + index number (survey questions)¶
In [ ]:
# Display the column names
print(colnames(data))
[1] "ID" [2] "Start time" [3] "Completion time" [4] "Email" [5] "Do you consider yourself a professional or a hobbyist in game development?" [6] "Primary area of work?" [7] "Years of experience in game development?" [8] "Are you in a lead role?" [9] "Team size?" [10] "Overall stance on Generative AI?2" [11] "Which areas do you think are less<U+00A0>" [12] "Art & Assets" [13] "Level Design" [14] "Storytelling" [15] "Sound Design" [16] "Voice Overs & Acting" [17] "Programming" [18] "Game Design" [19] "Marketing & PR" [20] "Music" [21] "Community management" [22] "Initial prototyping" [23] "Do you use Generative AI in your work?" [24] "Was it your own idea to begin using Generative AI or your employers?" [25] "Efficiency?" [26] "Quality?" [27] "Enjoyment?" [28] "Do you think of the ability to use Generative AI as an actual and useful competence in your area of work?" [29] "Do your coworkers use Generative AI in their work?" [30] "Do your peers in the industry use Generative AI in their work?" [31] "Do you disclose your use of Generative AI internally?" [32] "Do you disclose your use of Generative AI externally?" [33] "Do you perceive any stigma associated with the use of Generative AI internally?" [34] "Do you perceive any stigma associated with the use of Generative AI externally?" [35] "Would you care to elaborate?5" [36] "Would you care to elaborate?4" [37] "It will help shorten development timelines" [38] "It will lead to more individualized gaming experiences" [39] "It will impact staffing decisions" [40] "It will lead to smaller team sizes" [41] "It will democratize game development" [42] "It will make it cheaper to develop games" [43] "It will lead to lower quality games" [44] "It will lead to better games" [45] "What do you think will be the most promising innovations from Generative AI in game development?" [46] "What do you think will be the most negative consequences of Generative AI in game development?" [47] "Statement 1" [48] "Statement 23" [49] "Question4" [50] "What would be your ideal future Generative AI driven tool, that could help you in your area of work?" [51] "Creativity?" [52] "Anything else we should know / feedback?"
Syntax note¶
From here on we will use the index numbers to refer to the questions.
Example:
To return the contents of the fifth column "Do you consider yourself a professional or a hobbyist in game development?"
data[5] # Returns a data frame with the fifth column
data[[5]] # Returns the raw contents of the fifth column
Count of Professionals vs. Hobbyists¶
Here we'll access column [5] "Do you consider yourself a professional or a hobbyist in game development?"
In [ ]:
# Print the count of professionals vs. hobbyists
table(data[[5]])
# Plotting the count of professionals vs. hobbyists
ggplot(
as.data.frame(table(data[[5]])),
aes(x = Var1, y = Freq, fill = Var1)
) +
geom_bar(stat = "identity") +
labs(
title = "Professionals vs. Hobbyists",
x = "Type",
y = "Count"
)
Hobbyist Professional
200 107
Primary area of work¶
Here it's column [6] "Primary area of work?"
In [ ]:
# Show the distribution of 'Primary area of work?' in number of respondents
primary_area_count <- table(data[6])
# Censor abusive language in [6]
names(primary_area_count)[6] <- "Censored"
# Print the raw count of respondents in each primary area of work
primary_area_count
Art & Assets
35
Audio & Music
3
Auto body, lol
1
Defense
1
Design & Production
13
Censored
1
Generalist
10
Generalist (also solo developer)
125
I work a full-time job and do game dev as a hobby.
1
Marketing & PR
1
Project Lead, Programmer & Designer
1
Solo developer - I do everything
1
Technical & Programming
112
Travel & Customer Service
1
Whatever
1
Manual clean up of categories¶
In [ ]:
# Generalists
for (i in c(7, 9, 11, 12)) { # Combine the categories
primary_area_count[8] <- primary_area_count[8] + primary_area_count[i]
}
# Rename to Generalists
names(primary_area_count)[8] <- "Generalists"
# Other
for (i in c(3, 4, 6, 14)) { # Combine the categories
primary_area_count[15] <- primary_area_count[15] + primary_area_count[i]
}
# Rename to Other
names(primary_area_count)[15] <- "Other"
# Remove the categories that were combined
primary_area_count <- primary_area_count[-c(7, 9, 11, 12, 3, 4, 6, 14)]
# Output the cleaned up categories
primary_area_count
Art & Assets Audio & Music Design & Production
35 3 13
Generalists Marketing & PR Technical & Programming
138 1 112
Other
5
Visualizing the data¶
In [ ]:
par(bg = "white") # White background
# Distribution of primary areas of work
pie(
primary_area_count,
labels = paste(names(primary_area_count), primary_area_count),
col = rainbow(length(primary_area_count)),
main = "Distribution of Primary Areas of Work"
)
# And as a dotchart
dotchart(
primary_area_count,
labels = paste(names(primary_area_count), primary_area_count),
col = rainbow(length(primary_area_count)),
main = "Distribution of Primary Areas of Work"
)
primary_area_count
Art & Assets Audio & Music Design & Production
35 3 13
Generalists Marketing & PR Technical & Programming
138 1 112
Other
5
Distribution of Years of Experience¶
Here we're workin with column [7] "Years of experience in game development?"
In [ ]:
# Define the correct order for the levels
year_levels <- c(
"0-1 years",
"2-4 years",
"5-9 years",
"10-14 years",
"15-19 years",
"20 years or more"
)
# Convert the Years of experience column to a factor with specified levels
data[[7]] <- factor(data[[7]], levels = year_levels)
# Filter out 'Unknown' values
filtered_data <- data[!is.na(data[[7]]), ]
# Descriptive analysis: Distribution of years of experience
experience_distribution <- table(filtered_data[[7]])
# Plotting the distribution of years of experience
ggplot(
as.data.frame(experience_distribution),
aes(x = Var1, y = Freq, fill = Var1)
) +
geom_bar(stat = "identity") +
labs(
title = "Years of experience in game development?",
x = "Experience",
y = "Count"
)
Overall Stance on Generative AI¶
In [ ]:
# Define the correct order for the levels
stance_levels <- c(
"Very negative",
"Negative",
"Neutral",
"Positive",
"Very positive"
)
# Convert the 'Overall stance on Generative AI?2'
# column to a factor with specified levels
data[[10]] <- factor(data[[10]], levels = stance_levels)
# Exclude 'Unknown' values
filtered_data <- data[!is.na(data[[10]]), ]
# Descriptive analysis: Overall stance on Generative AI
stance_distribution <- table(filtered_data[[10]])
# Plotting the overall stance on Generative AI
ggplot(
as.data.frame(stance_distribution),
aes(x = Var1, y = Freq, fill = Var1)
) +
geom_bar(stat = "identity") +
labs(title = "Overall Stance on Generative AI", x = "Stance", y = "Count")
Correlation Analysis¶
In [ ]:
# Convert categorical data to numeric for correlation analysis
data_numeric <- data %>%
mutate(
YearsExperience = case_when(
.[[7]] == "0-1 years" ~ 1,
.[[7]] == "2-4 years" ~ 2,
.[[7]] == "5-9 years" ~ 3,
.[[7]] == "10-14 years" ~ 4,
.[[7]] == "15-19 years" ~ 5,
.[[7]] == "20 years or more" ~ 6,
TRUE ~ 0
),
StanceOnGenAI = case_when(
.[[10]] == "Very positive" ~ 5,
.[[10]] == "Positive" ~ 4,
.[[10]] == "Neutral" ~ 3,
.[[10]] == "Negative" ~ 2,
.[[10]] == "Very negative" ~ 1,
TRUE ~ 0
)
)
# Calculate correlation
correlation <- cor(
data_numeric$YearsExperience,
data_numeric$StanceOnGenAI,
method = "spearman",
use = "complete.obs"
)
print(correlation)
[1] -0.08107992
In [ ]:
# Definer niveauerne for erfaring i industrien
year_levels <- c(
"0-1 years",
"2-4 years",
"5-9 years",
"10-14 years",
"15-19 years",
"20 years or more"
)
# Definer niveauerne for holdninger til generativ AI
stance_levels <- c(
"Very negative",
"Negative",
"Neutral",
"Positive",
"Very positive"
)
# Konverter kolonnerne til passende formater
data <- data %>%
mutate(
YearsExperience = factor(`Years of experience in game development?`, levels = year_levels),
StanceOnGenAI = case_when(
`Overall stance on Generative AI?2` == "Very positive" ~ 5,
`Overall stance on Generative AI?2` == "Positive" ~ 4,
`Overall stance on Generative AI?2` == "Neutral" ~ 3,
`Overall stance on Generative AI?2` == "Negative" ~ 2,
`Overall stance on Generative AI?2` == "Very negative" ~ 1,
TRUE ~ 0
)
)
# Filtrer data for at fjerne 'Unknown' værdier
filtered_data <- data %>%
filter(!is.na(YearsExperience) & StanceOnGenAI != 0)
# Udfør ANOVA-testen
anova_result <- aov(StanceOnGenAI ~ YearsExperience, data = filtered_data)
# Opsummering af ANOVA-resultaterne
summary(anova_result)
# Udfør Kruskal-Wallis test som en ikke-parametrisk alternativ
kruskal_result <- kruskal.test(StanceOnGenAI ~ YearsExperience, data = filtered_data)
# Opsummering af Kruskal-Wallis test resultater
print(kruskal_result)
Df Sum Sq Mean Sq F value Pr(>F) YearsExperience 5 4.11 0.823 0.397 0.85 Residuals 101 209.60 2.075
Kruskal-Wallis rank sum test data: StanceOnGenAI by YearsExperience Kruskal-Wallis chi-squared = 1.8417, df = 5, p-value = 0.8706
Sentiment Analysis¶
Wordcloud¶
"What do you think will be the most promising innovations from Generative AI in game development?"¶
In [ ]:
# Preprocess the text data
data[[45]] <- iconv(data[[45]], "latin1", "UTF-8") # Convert to UTF-8
data[[45]] <- tolower(data[[45]]) # Convert to lowercase
data[[45]] <- removePunctuation(data[[45]]) # Remove punctuation
data[[45]] <- removeNumbers(data[[45]]) # Remove numbers
data[[45]] <- removeWords(data[[45]], stopwords("smart")) # Remove stopwords
data[[45]] <- gsub("\\bgame\\b", "games", data[[45]]) # turn game into games
data[[45]] <- stripWhitespace(data[[45]]) # Remove extra whitespaces
par(bg = "white") # White background
# Create a wordcloud
wordcloud(
data[[45]],
colors = rainbow(10),
random.order = FALSE,
scale = c(5, 0.5),
max.words = 250
)
# Add a title to the wordcloud
title(main = "Most promising innovations from GenAI?")
# now withouth "games"
data[[45]] <- removeWords(data[[45]], "games")
wordcloud(
data[[45]],
colors = rainbow(10),
random.order = FALSE,
scale = c(4, 0.5),
max.words = 250
)
# Add a title to the wordcloud
title(main = "Most promising innovations from GenAI? (-games)")
"What do you think will be the most negative consequences of Generative AI in game development?"¶
In [ ]:
# wordcloud
# Preprocess the text data
data[[46]] <- iconv(data[[46]], "latin1", "UTF-8") # Convert to UTF-8
data[[46]] <- tolower(data[[46]]) # Convert to lowercase
data[[46]] <- removePunctuation(data[[46]]) # Remove punctuation
data[[46]] <- removeNumbers(data[[46]]) # Remove numbers
data[[46]] <- removeWords(data[[46]], stopwords("smart")) # Remove stopwords
data[[46]] <- gsub("\\bgame\\b", "games", data[[46]]) # turn game into games
data[[46]] <- stripWhitespace(data[[46]]) # Remove extra whitespaces
par(bg = "white") # White background
# Create a wordcloud
wordcloud(
data[[46]],
colors = rainbow(10),
random.order = FALSE,
scale = c(6, 0.5),
max.words = 250
)
# Add a title to the wordcloud
title(main = "Most negative consequences from GenAI?")
# now withouth "games"
data[[46]] <- removeWords(data[[46]], "games")
wordcloud(
data[[46]],
colors = rainbow(10),
random.order = FALSE,
scale = c(6, 0.5),
max.words = 250
)
# Add a title to the wordcloud
title(main = "Most negative consequences from GenAI? (-games)")
In [ ]:
# Extract open-ended responses
open_ended_responses <- data[[45]]
# Perform sentiment analysis
sentiment_scores_0 <- get_nrc_sentiment(open_ended_responses)
# Summarize sentiment scores
sentiment_summary_0 <- colSums(sentiment_scores_0)
print(sentiment_summary_0)
par(bg = "white") # White background
# Plot sentiment scores
barplot(
sentiment_summary_0,
las = 2,
col = rainbow(10),
main = "Sentiment Analysis of most promising inovations from genAI?"
)
anger anticipation disgust fear joy sadness
23 87 22 28 85 48
surprise trust negative positive
63 100 66 234
In [ ]:
# Extract open-ended responses
open_ended_responses <- data[[46]]
# Perform sentiment analysis
sentiment_scores_1 <- get_nrc_sentiment(open_ended_responses)
# Summarize sentiment scores
sentiment_summary_1 <- colSums(sentiment_scores_1)
print(sentiment_summary_1)
par(bg = "white") # White background
# Plot sentiment scores
barplot(
sentiment_summary_1,
las = 2,
col = rainbow(10),
main = "Sentiment Analysis of most negative consequences from genAI?"
)
anger anticipation disgust fear joy sadness
89 143 71 107 119 147
surprise trust negative positive
83 137 226 329
In [ ]:
# Extract open-ended responses
open_ended_responses <- data[[50]]
# Perform sentiment analysis
sentiment_scores_2 <- get_nrc_sentiment(open_ended_responses)
# Summarize sentiment scores
sentiment_summary_2 <- colSums(sentiment_scores_2)
print(sentiment_summary_2)
par(bg = "white") # White background
# Plot sentiment scores
barplot(
sentiment_summary_2,
las = 2,
col = rainbow(10),
main = "Sentiment Analysis of ideal future genAI driven tool?"
)
anger anticipation disgust fear joy sadness
24 96 25 36 94 56
surprise trust negative positive
42 117 75 276
In [ ]:
# Extract open-ended responses
open_ended_responses <- data[[52]]
# Perform sentiment analysis
sentiment_scores_3 <- get_nrc_sentiment(open_ended_responses)
# Summarize sentiment scores
sentiment_summary_3 <- colSums(sentiment_scores_3)
print(sentiment_summary_3)
par(bg = "white") # White background
# Plot sentiment scores
barplot(
sentiment_summary_3,
las = 2,
col = rainbow(10),
main = "Sentiment Analysis of 'Anything else we should know?'"
)
anger anticipation disgust fear joy sadness
23 59 26 32 43 44
surprise trust negative positive
29 74 58 145
Group Comparisons¶
In [ ]:
# Define the correct order for the levels
stance_levels <- c(
"Very negative",
"Negative",
"Neutral",
"Positive",
"Very positive"
)
# Convert the 'Overall stance on GenAI' column to a factor with specified levels
data[[10]] <- factor(data[[10]], levels = stance_levels)
# Convert categorical data to numeric for stance on Generative AI
data <- data %>%
mutate(
StanceOnGenAI = case_when(
.[[10]] == "Very positive" ~ 5,
.[[10]] == "Positive" ~ 4,
.[[10]] == "Neutral" ~ 3,
.[[10]] == "Negative" ~ 2,
.[[10]] == "Very negative" ~ 1,
TRUE ~ NA_real_
)
)
# Filter data for professionals and hobbyists
professionals <- data %>% filter(data[[5]] == "Professional")
hobbyists <- data %>% filter(data[[5]] == "Hobbyist")
# Calculate average stance on Generative AI
avg_stance_professionals <- mean(professionals$StanceOnGenAI, na.rm = TRUE)
avg_stance_hobbyists <- mean(hobbyists$StanceOnGenAI, na.rm = TRUE)
print(
paste(
"Average stance on GenAI for professionals:",
avg_stance_professionals
)
)
print(
paste(
"Average stance on GenAI for hobbyists:",
avg_stance_hobbyists
)
)
# Perform a t-test to compare the average stance
# on Generative AI between professionals and hobbyists
t_test_result <- t.test(
professionals$StanceOnGenAI,
hobbyists$StanceOnGenAI,
alternative = "two.sided",
mu = 0,
conf.level = 0.95
)
# Print the t-test results
print(t_test_result)
# visualize the t-test results
ggplot() +
geom_density(
aes(x = professionals$StanceOnGenAI, fill = "Professionals"),
alpha = 0.5
) +
geom_density(
aes(x = hobbyists$StanceOnGenAI, fill = "Hobbyists"),
alpha = 0.5
) +
geom_vline(
xintercept = t_test_result$estimate,
) +
labs(
title = "Distribution of Stance on Generative AI",
x = "Stance on Generative AI",
y = "Density"
)
[1] "Average stance on GenAI for professionals: 2.5981308411215" [1] "Average stance on GenAI for hobbyists: 2.795" Welch Two Sample t-test data: professionals$StanceOnGenAI and hobbyists$StanceOnGenAI t = -1.1712, df = 210.38, p-value = 0.2428 alternative hypothesis: true difference in means is not equal to 0 95 percent confidence interval: -0.5282309 0.1344926 sample estimates: mean of x mean of y 2.598131 2.795000
Stance on Generative AI vs. Years of Experience¶
We'll be working with [7] "Years of experience in game development?" and [10] "Overall stance on Generative AI?2"
In [ ]:
# Factorize the 'Years of experience' and 'Stance on GenAI' columns
data <- data %>%
mutate(
ExperienceGroup = factor(data[[7]]),
StanceOnGenAI = as.numeric(data[[10]])
)
In [ ]:
# Boxplot for Stance on GenAI vs. Experience Group
ggplot(
data, aes(
x = ExperienceGroup,
y = StanceOnGenAI,
fill = ExperienceGroup
)
) +
geom_boxplot() +
labs(
title = "Stance on Generative AI vs. Years of Experience",
x = "Years of Experience",
y = "Stance on Generative AI"
)
In [ ]:
# ANOVA test for Stance on GenAI vs. Experience Group
anova_result <- aov(StanceOnGenAI ~ ExperienceGroup, data = data)
summary(anova_result)
Df Sum Sq Mean Sq F value Pr(>F) ExperienceGroup 5 4.11 0.823 0.397 0.85 Residuals 101 209.60 2.075 200 observations deleted due to missingness
Main area of work VS Stance on GenAI¶
Here it's [6] "Primary area of work?" and [10] "Overall stance on Generative AI?2"
In [ ]:
# Konverter relevante kolonner til faktorer
data <- data %>%
mutate(
UsesGenAI = factor(`Do you use Generative AI in your work?`),
StanceOnGenAI = as.numeric(`Overall stance on Generative AI?2`)
)
# Boxplot for Stance on GenAI vs. Uses GenAI
ggplot(data, aes(x = UsesGenAI, y = StanceOnGenAI, fill = UsesGenAI)) +
geom_boxplot() +
labs(
title = "Stance on Generative AI vs. Uses Generative AI",
x = "Uses Generative AI",
y = "Stance on Generative AI"
)
# T-test for Stance on GenAI vs. Uses GenAI
t_test_result <- t.test(StanceOnGenAI ~ UsesGenAI, data = data)
print(t_test_result)
Welch Two Sample t-test
data: StanceOnGenAI by UsesGenAI
t = -12.196, df = 251.95, p-value < 2.2e-16
alternative hypothesis: true difference in means between group No and group Yes is not equal to 0
95 percent confidence interval:
-1.891231 -1.365338
sample estimates:
mean in group No mean in group Yes
2.100529 3.728814
Users vs non-users of GenAI in different areas¶
In [ ]:
# Define the levels for the factor
agree_levels <- c(
"Strongly disagree",
"Disagree",
"Neutral",
"Agree",
"Strongly agree"
)
In [ ]:
# Convert relevant columns to numeric
data <- data %>%
mutate(
UsesGenAI = factor(`Do you use Generative AI in your work?`),
ShortenTimelines = as.numeric(factor(
`It will help shorten development timelines`,
agree_levels
)),
IndividualizedExperiences = as.numeric(factor(
`It will lead to more individualized gaming experiences`,
agree_levels
)),
ImpactStaffing = as.numeric(factor(
`It will impact staffing decisions`,
agree_levels
)),
SmallerTeams = as.numeric(factor(
`It will lead to smaller team sizes`,
agree_levels
)),
DemocratizeDevelopment = as.numeric(factor(
`It will democratize game development`,
agree_levels
)),
CheaperDevelopment = as.numeric(factor(
`It will make it cheaper to develop games`,
agree_levels
)),
LowerQuality = as.numeric(factor(
`It will lead to lower quality games`,
agree_levels
)),
BetterGames = as.numeric(factor(
`It will lead to better games`,
agree_levels
))
)
# Calculate mean responses for each question grouped by UsesGenAI
mean_responses <- data %>%
group_by(UsesGenAI) %>%
summarise(
MeanShortenTimelines = mean(
ShortenTimelines,
na.rm = TRUE
),
MeanIndividualizedExperiences = mean(
IndividualizedExperiences,
na.rm = TRUE
),
MeanImpactStaffing = mean(
ImpactStaffing,
na.rm = TRUE
),
MeanSmallerTeams = mean(
SmallerTeams,
na.rm = TRUE
),
MeanDemocratizeDevelopment = mean(
DemocratizeDevelopment,
na.rm = TRUE
),
MeanCheaperDevelopment = mean(
CheaperDevelopment,
na.rm = TRUE
),
MeanLowerQuality = mean(
LowerQuality,
na.rm = TRUE
),
MeanBetterGames = mean(
BetterGames,
na.rm = TRUE
)
)
mean_responses
# Pivot the data to a long format for plotting
data_long <- data %>%
pivot_longer(
cols = starts_with("ShortenTimelines"):starts_with("BetterGames"),
names_to = "Question",
values_to = "Response"
)
# Improve the plot layout and labels
ggplot(data_long, aes(x = UsesGenAI, y = Response, fill = UsesGenAI)) +
geom_boxplot() +
facet_wrap(~Question, scales = "free_y", ncol = 2) +
labs(
title = "Comparison of Perceptions by Generative AI Usage",
x = "Uses Generative AI",
y = "Response (1=Strongly Disagree, 5=Strongly Agree)",
fill = "Uses Generative AI"
)
| UsesGenAI | MeanShortenTimelines | MeanIndividualizedExperiences | MeanImpactStaffing | MeanSmallerTeams | MeanDemocratizeDevelopment | MeanCheaperDevelopment | MeanLowerQuality | MeanBetterGames |
|---|---|---|---|---|---|---|---|---|
| <fct> | <dbl> | <dbl> | <dbl> | <dbl> | <dbl> | <dbl> | <dbl> | <dbl> |
| No | 2.851852 | 2.137566 | 4.121693 | 3.328042 | 1.746032 | 2.671958 | 4.417989 | 1.809524 |
| Yes | 4.008475 | 3.364407 | 3.991525 | 3.508475 | 3.076271 | 3.635593 | 3.652542 | 3.127119 |
In [ ]:
# Perform t-tests for each question
t_test_results <- list()
questions <- c(
"ShortenTimelines",
"IndividualizedExperiences",
"ImpactStaffing",
"SmallerTeams",
"DemocratizeDevelopment",
"CheaperDevelopment",
"LowerQuality",
"BetterGames"
)
for (question in questions) {
t_test_result <- t.test(get(question) ~ UsesGenAI, data = data)
t_test_results[[question]] <- t_test_result
}
# Print t-test results
for (question in questions) {
print(paste("T-test for", question))
print(t_test_results[[question]])
}
[1] "T-test for ShortenTimelines"
Welch Two Sample t-test
data: get(question) by UsesGenAI
t = -9.3595, df = 295.77, p-value < 2.2e-16
alternative hypothesis: true difference in means between group No and group Yes is not equal to 0
95 percent confidence interval:
-1.3998252 -0.9134202
sample estimates:
mean in group No mean in group Yes
2.851852 4.008475
[1] "T-test for IndividualizedExperiences"
Welch Two Sample t-test
data: get(question) by UsesGenAI
t = -8.3172, df = 253.16, p-value = 5.597e-15
alternative hypothesis: true difference in means between group No and group Yes is not equal to 0
95 percent confidence interval:
-1.5173360 -0.9363453
sample estimates:
mean in group No mean in group Yes
2.137566 3.364407
[1] "T-test for ImpactStaffing"
Welch Two Sample t-test
data: get(question) by UsesGenAI
t = 1.1843, df = 278.51, p-value = 0.2373
alternative hypothesis: true difference in means between group No and group Yes is not equal to 0
95 percent confidence interval:
-0.0861925 0.3465279
sample estimates:
mean in group No mean in group Yes
4.121693 3.991525
[1] "T-test for SmallerTeams"
Welch Two Sample t-test
data: get(question) by UsesGenAI
t = -1.3938, df = 278.46, p-value = 0.1645
alternative hypothesis: true difference in means between group No and group Yes is not equal to 0
95 percent confidence interval:
-0.43527137 0.07440687
sample estimates:
mean in group No mean in group Yes
3.328042 3.508475
[1] "T-test for DemocratizeDevelopment"
Welch Two Sample t-test
data: get(question) by UsesGenAI
t = -9.1581, df = 194.75, p-value < 2.2e-16
alternative hypothesis: true difference in means between group No and group Yes is not equal to 0
95 percent confidence interval:
-1.616711 -1.043767
sample estimates:
mean in group No mean in group Yes
1.746032 3.076271
[1] "T-test for CheaperDevelopment"
Welch Two Sample t-test
data: get(question) by UsesGenAI
t = -6.4659, df = 260.41, p-value = 4.944e-10
alternative hypothesis: true difference in means between group No and group Yes is not equal to 0
95 percent confidence interval:
-1.2570992 -0.6701719
sample estimates:
mean in group No mean in group Yes
2.671958 3.635593
[1] "T-test for LowerQuality"
Welch Two Sample t-test
data: get(question) by UsesGenAI
t = 6.0497, df = 225.21, p-value = 5.988e-09
alternative hypothesis: true difference in means between group No and group Yes is not equal to 0
95 percent confidence interval:
0.5161216 1.0147725
sample estimates:
mean in group No mean in group Yes
4.417989 3.652542
[1] "T-test for BetterGames"
Welch Two Sample t-test
data: get(question) by UsesGenAI
t = -9.8995, df = 237.71, p-value < 2.2e-16
alternative hypothesis: true difference in means between group No and group Yes is not equal to 0
95 percent confidence interval:
-1.579796 -1.055393
sample estimates:
mean in group No mean in group Yes
1.809524 3.127119
In [ ]:
# Custom function to convert columns to numeric factor levels
convert_to_numeric <- function(column, levels) {
as.numeric(factor(column, levels = levels))
}
# Convert relevant columns to factors and numeric
data <- data %>%
mutate(
LeadRole = ifelse(
is.na(`Are you in a lead role?`),
"Hobbyist", as.character(`Are you in a lead role?`)
),
LeadRole = factor(LeadRole, levels = c("No", "Yes", "Hobbyist")),
ShortenTimelines = convert_to_numeric(
`It will help shorten development timelines`,
agree_levels
),
IndividualizedExperiences = convert_to_numeric(
`It will lead to more individualized gaming experiences`,
agree_levels
),
ImpactStaffing = convert_to_numeric(
`It will impact staffing decisions`,
agree_levels
),
SmallerTeams = convert_to_numeric(
`It will lead to smaller team sizes`,
agree_levels
),
DemocratizeDevelopment = convert_to_numeric(
`It will democratize game development`,
agree_levels
),
CheaperDevelopment = convert_to_numeric(
`It will make it cheaper to develop games`,
agree_levels
),
LowerQuality = convert_to_numeric(
`It will lead to lower quality games`,
agree_levels
),
BetterGames = convert_to_numeric(
`It will lead to better games`,
agree_levels
)
)
# Calculate mean responses for each question grouped by LeadRole
mean_responses_lead <- data %>%
group_by(LeadRole) %>%
summarise(
MeanShortenTimelines = mean(
ShortenTimelines,
na.rm = TRUE
),
MeanIndividualizedExperiences = mean(
IndividualizedExperiences,
na.rm = TRUE
),
MeanImpactStaffing = mean(
ImpactStaffing,
na.rm = TRUE
),
MeanSmallerTeams = mean(
SmallerTeams,
na.rm = TRUE
),
MeanDemocratizeDevelopment = mean(
DemocratizeDevelopment,
na.rm = TRUE
),
MeanCheaperDevelopment = mean(
CheaperDevelopment,
na.rm = TRUE
),
MeanLowerQuality = mean(
LowerQuality,
na.rm = TRUE
),
MeanBetterGames = mean(
BetterGames,
na.rm = TRUE
)
)
print(mean_responses_lead)
# Pivot the data to a long format for plotting
data_long_lead <- data %>%
pivot_longer(
cols = ShortenTimelines:BetterGames,
names_to = "Question",
values_to = "Response"
)
# Plot each question in a facet grid
ggplot(data_long_lead, aes(x = LeadRole, y = Response, fill = LeadRole)) +
geom_boxplot() +
facet_wrap(~Question, scales = "free_y", ncol = 2) +
labs(
title = "Comparison of Perceptions by Lead Role",
x = "Lead Role",
y = "Response (1=Strongly Disagree, 5=Strongly Agree)",
fill = "Lead Role"
)
# A tibble: 3 x 9 LeadRole MeanShortenTimelines MeanIndividualizedExperiences MeanImpactStaffing <fct> <dbl> <dbl> <dbl> 1 No 3.36 2.89 4.20 2 Yes 3.14 2.32 3.86 3 Hobbyist 3.33 2.64 4.11 # i 5 more variables: MeanSmallerTeams <dbl>, MeanDemocratizeDevelopment <dbl>, # MeanCheaperDevelopment <dbl>, MeanLowerQuality <dbl>, MeanBetterGames <dbl>
In [ ]:
# Filter out hobbyists for the t-tests
data_filtered <- data %>%
filter(LeadRole %in% c("No", "Yes"))
# Perform t-tests for each question
t_test_results <- list()
questions <- c(
"ShortenTimelines",
"IndividualizedExperiences",
"ImpactStaffing",
"SmallerTeams",
"DemocratizeDevelopment",
"CheaperDevelopment",
"LowerQuality",
"BetterGames"
)
for (question in questions) {
t_test_result <- t.test(get(question) ~ LeadRole, data = data_filtered)
t_test_results[[question]] <- t_test_result
}
# Print t-test results
for (question in questions) {
print(paste("T-test for", question))
print(t_test_results[[question]])
}
[1] "T-test for ShortenTimelines"
Welch Two Sample t-test
data: get(question) by LeadRole
t = 0.90523, df = 92.772, p-value = 0.3677
alternative hypothesis: true difference in means between group No and group Yes is not equal to 0
95 percent confidence interval:
-0.2635614 0.7051199
sample estimates:
mean in group No mean in group Yes
3.363636 3.142857
[1] "T-test for IndividualizedExperiences"
Welch Two Sample t-test
data: get(question) by LeadRole
t = 2.2615, df = 90.679, p-value = 0.02612
alternative hypothesis: true difference in means between group No and group Yes is not equal to 0
95 percent confidence interval:
0.06918259 1.06862405
sample estimates:
mean in group No mean in group Yes
2.886364 2.317460
[1] "T-test for ImpactStaffing"
Welch Two Sample t-test
data: get(question) by LeadRole
t = 1.7029, df = 101.76, p-value = 0.09163
alternative hypothesis: true difference in means between group No and group Yes is not equal to 0
95 percent confidence interval:
-0.05724882 0.75205402
sample estimates:
mean in group No mean in group Yes
4.204545 3.857143
[1] "T-test for SmallerTeams"
Welch Two Sample t-test
data: get(question) by LeadRole
t = 0.57107, df = 87.702, p-value = 0.5694
alternative hypothesis: true difference in means between group No and group Yes is not equal to 0
95 percent confidence interval:
-0.3453515 0.6238507
sample estimates:
mean in group No mean in group Yes
3.409091 3.269841
[1] "T-test for DemocratizeDevelopment"
Welch Two Sample t-test
data: get(question) by LeadRole
t = -0.63141, df = 97.143, p-value = 0.5293
alternative hypothesis: true difference in means between group No and group Yes is not equal to 0
95 percent confidence interval:
-0.6486929 0.3355615
sample estimates:
mean in group No mean in group Yes
1.954545 2.111111
[1] "T-test for CheaperDevelopment"
Welch Two Sample t-test
data: get(question) by LeadRole
t = 0.35169, df = 94.569, p-value = 0.7259
alternative hypothesis: true difference in means between group No and group Yes is not equal to 0
95 percent confidence interval:
-0.4323495 0.6184967
sample estimates:
mean in group No mean in group Yes
3.045455 2.952381
[1] "T-test for LowerQuality"
Welch Two Sample t-test
data: get(question) by LeadRole
t = -0.67785, df = 91.141, p-value = 0.4996
alternative hypothesis: true difference in means between group No and group Yes is not equal to 0
95 percent confidence interval:
-0.5784907 0.2841184
sample estimates:
mean in group No mean in group Yes
4.090909 4.238095
[1] "T-test for BetterGames"
Welch Two Sample t-test
data: get(question) by LeadRole
t = 0.5885, df = 89.019, p-value = 0.5577
alternative hypothesis: true difference in means between group No and group Yes is not equal to 0
95 percent confidence interval:
-0.3351935 0.6173003
sample estimates:
mean in group No mean in group Yes
2.204545 2.063492
In [ ]:
# Convert relevant columns to factors and numeric, and filter out hobbyists
data <- data %>%
mutate(
TeamSize = factor(
`Team size?`,
levels = c(
"Solo",
"2-9",
"10-49",
"50-149",
"150-299",
"300+"
)
),
ShortenTimelines = as.numeric(factor(
`It will help shorten development timelines`,
agree_levels
)),
IndividualizedExperiences = as.numeric(factor(
`It will lead to more individualized gaming experiences`,
agree_levels
)),
ImpactStaffing = as.numeric(factor(
`It will impact staffing decisions`,
agree_levels
)),
SmallerTeams = as.numeric(factor(
`It will lead to smaller team sizes`,
agree_levels
)),
DemocratizeDevelopment = as.numeric(factor(
`It will democratize game development`,
agree_levels
)),
CheaperDevelopment = as.numeric(factor(
`It will make it cheaper to develop games`,
agree_levels
)),
LowerQuality = as.numeric(factor(
`It will lead to lower quality games`,
agree_levels
)),
BetterGames = as.numeric(factor(
`It will lead to better games`,
agree_levels
))
) %>%
filter(!is.na(TeamSize))
# Perform ANOVA for each question
anova_results <- list()
questions <- c(
"ShortenTimelines",
"IndividualizedExperiences",
"ImpactStaffing",
"SmallerTeams",
"DemocratizeDevelopment",
"CheaperDevelopment",
"LowerQuality",
"BetterGames"
)
for (question in questions) {
anova_result <- aov(get(question) ~ TeamSize, data = data)
anova_results[[question]] <- summary(anova_result)
}
# Print ANOVA results
for (question in questions) {
print(paste("ANOVA for", question))
print(anova_results[[question]])
}
# Pivot the data to a long format for plotting
data_long_team <- data %>%
pivot_longer(
cols = starts_with("ShortenTimelines"):starts_with("BetterGames"),
names_to = "Question",
values_to = "Response"
)
# Plot each question in a facet grid
ggplot(data_long_team, aes(x = TeamSize, y = Response, fill = TeamSize)) +
geom_boxplot() +
facet_wrap(~Question, scales = "free_y", ncol = 2) +
labs(
title = "Comparison of Perceptions by Team Size",
x = "Team Size",
y = "Response (1=Strongly Disagree, 5=Strongly Agree)",
fill = "Team Size"
)
[1] "ANOVA for ShortenTimelines"
Df Sum Sq Mean Sq F value Pr(>F)
TeamSize 5 7.83 1.566 1.019 0.411
Residuals 101 155.33 1.538
[1] "ANOVA for IndividualizedExperiences"
Df Sum Sq Mean Sq F value Pr(>F)
TeamSize 5 12.56 2.512 1.529 0.187
Residuals 101 165.91 1.643
[1] "ANOVA for ImpactStaffing"
Df Sum Sq Mean Sq F value Pr(>F)
TeamSize 5 9.38 1.876 1.653 0.153
Residuals 101 114.62 1.135
[1] "ANOVA for SmallerTeams"
Df Sum Sq Mean Sq F value Pr(>F)
TeamSize 5 12.69 2.538 1.769 0.126
Residuals 101 144.86 1.434
[1] "ANOVA for DemocratizeDevelopment"
Df Sum Sq Mean Sq F value Pr(>F)
TeamSize 5 9.87 1.975 1.225 0.303
Residuals 101 162.89 1.613
[1] "ANOVA for CheaperDevelopment"
Df Sum Sq Mean Sq F value Pr(>F)
TeamSize 5 6.94 1.389 0.754 0.585
Residuals 101 186.05 1.842
[1] "ANOVA for LowerQuality"
Df Sum Sq Mean Sq F value Pr(>F)
TeamSize 5 21.07 4.214 3.994 0.00239 **
Residuals 101 106.56 1.055
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
[1] "ANOVA for BetterGames"
Df Sum Sq Mean Sq F value Pr(>F)
TeamSize 5 13.26 2.652 1.911 0.099 .
Residuals 101 140.16 1.388
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
How useful GenAI are in different areas¶
Prepare data¶
In [ ]:
# Define usefulness columns
usefulness_columns <- c(
"Art & Assets",
"Level Design",
"Storytelling",
"Sound Design",
"Voice Overs & Acting",
"Programming",
"Game Design",
"Marketing & PR",
"Music",
"Community management",
"Initial prototyping"
)
# Convert usefulness columns to factors with specific levels
data <- data %>%
mutate(across(
all_of(usefulness_columns),
~ factor(., levels = c(
"Not useful",
"Somewhat useful",
"Very useful"
))
))
# Map categorical responses to numeric values
usefulness_mapping <- c(
"Not useful" = 1,
"Somewhat useful" = 2,
"Very useful" = 3
)
data <- data %>%
mutate(across(
all_of(usefulness_columns),
~ usefulness_mapping[as.character(.)]
))
Calculate Average Usefulness Scores¶
In [ ]:
# Calculate mean usefulness scores for each area using updated syntax
mean_usefulness_scores <- data %>%
summarise(across(
all_of(usefulness_columns),
\(x) mean(x, na.rm = TRUE)
)) %>%
pivot_longer(
cols = everything(),
names_to = "Area",
values_to = "MeanScore"
)
# Print mean usefulness scores to check for anomalies
print(mean_usefulness_scores)
# A tibble: 11 x 2 Area MeanScore <chr> <dbl> 1 Art & Assets 1.83 2 Level Design 1.39 3 Storytelling 1.52 4 Sound Design 1.44 5 Voice Overs & Acting 1.66 6 Programming 1.93 7 Game Design 1.28 8 Marketing & PR 1.51 9 Music 1.35 10 Community management 1.35 11 Initial prototyping 1.91
Visualization¶
In [ ]:
# Bar plot of average usefulness scores with y-axis fixed
ggplot(
mean_usefulness_scores,
aes(x = reorder(Area, -MeanScore), y = MeanScore, fill = Area)
) +
geom_bar(stat = "identity") +
labs(
title = "Average Perceived Usefulness of Generative AI in Different Areas",
x = "Area",
y = "Mean Usefulness Score (1=Not useful, 2=Somewhat useful, 3=Very useful)"
) +
theme(axis.text.x = element_text(angle = 45, hjust = 1))
Perceived Usefulness of Generative AI in Different Areas by Lead Role¶
In [ ]:
# Ensure the lead role column is correctly labeled
data <- data %>%
mutate(LeadRole = factor(
data[[8]],
levels = c("Yes", "No")
))
# Remove rows with all NA values in usefulness columns
data <- data %>%
filter(rowSums(!is.na(select(., all_of(usefulness_columns)))) > 0)
# Pivot the data to a long format for ggplot2
usefulness_long <- data %>%
pivot_longer(
cols = all_of(usefulness_columns),
names_to = "Area",
values_to = "UsefulnessScore"
)
# Remove rows with NA values in UsefulnessScore or LeadRole
usefulness_long <- usefulness_long %>%
filter(
!is.na(UsefulnessScore),
!is.na(LeadRole)
)
# Calculate mean and standard error for each group
usefulness_summary <- usefulness_long %>%
group_by(Area, LeadRole) %>%
summarise(
MeanScore = mean(UsefulnessScore, na.rm = TRUE),
SE = sd(UsefulnessScore, na.rm = TRUE) / sqrt(n())
)
# Bar plot with error bars side by side
ggplot(usefulness_summary, aes(
x = Area,
y = MeanScore,
fill = LeadRole
)) +
geom_bar(
stat = "identity",
position = position_dodge(width = 0.9)
) +
geom_errorbar(
aes(
ymin = MeanScore - SE,
ymax = MeanScore + SE
),
width = 0.2,
position = position_dodge(width = 0.9)
) +
labs(
title = "Perceived Usefulness of Generative AI
in Different Areas by Lead Role",
x = "Area",
y = "Mean Usefulness Score:
1 = Not useful
2 = Somewhat useful
3 = Very useful"
) +
theme(axis.text.x = element_text(angle = 45, hjust = 1))
`summarise()` has grouped output by 'Area'. You can override using the `.groups` argument.
Perceived Usefulness of Generative AI in Different Areas by Professional or hobbyist¶
In [ ]:
# Reset to source data
data <- read_excel(file_path)
data <- data %>%
mutate(across(
all_of(usefulness_columns),
~ usefulness_mapping[as.character(.)]
))
# Ensure the role column is correctly labeled
data <- data %>%
mutate(Role = factor(
data[[5]],
levels = c("Professional", "Hobbyist")
))
# Remove rows with all NA values in usefulness columns
data <- data %>%
filter(rowSums(!is.na(select(., all_of(usefulness_columns)))) > 0)
# Pivot the data to a long format for ggplot2
usefulness_long <- data %>%
pivot_longer(
cols = all_of(usefulness_columns),
names_to = "Area",
values_to = "UsefulnessScore"
)
# Remove rows with NA values in UsefulnessScore or Role
usefulness_long <- usefulness_long %>%
filter(!is.na(UsefulnessScore), !is.na(Role))
# Calculate mean and standard error for each group
usefulness_summary <- usefulness_long %>%
group_by(Area, Role) %>%
summarise(
MeanScore = mean(UsefulnessScore, na.rm = TRUE),
SE = sd(UsefulnessScore, na.rm = TRUE) / sqrt(n()),
.groups = "drop" # Explicitly drop the grouping
)
# Bar plot with error bars side by side
ggplot(usefulness_summary, aes(
x = Area,
y = MeanScore,
fill = Role
)) +
geom_bar(
stat = "identity",
position = position_dodge(width = 0.9)
) +
geom_errorbar(
aes(
ymin = MeanScore - SE,
ymax = MeanScore + SE
),
width = 0.2,
position = position_dodge(width = 0.9)
) +
labs(
title = "Perceived Usefulness of Generative AI
in Different Areas by Role",
x = "Area",
y = "Mean Usefulness Score
1 = Not useful
2 = Somewhat useful
3 = Very useful"
) +
theme(axis.text.x = element_text(angle = 45, hjust = 1))
# Print summary
usefulness_summary
| Area | Role | MeanScore | SE |
|---|---|---|---|
| <chr> | <fct> | <dbl> | <dbl> |
| Art & Assets | Professional | 1.831776 | 0.06961868 |
| Art & Assets | Hobbyist | 1.945000 | 0.05314127 |
| Community management | Professional | 1.345794 | 0.05953858 |
| Community management | Hobbyist | 1.270000 | 0.03928379 |
| Game Design | Professional | 1.280374 | 0.05107622 |
| Game Design | Hobbyist | 1.305000 | 0.03959436 |
| Initial prototyping | Professional | 1.906542 | 0.07456771 |
| Initial prototyping | Hobbyist | 1.930000 | 0.05559636 |
| Level Design | Professional | 1.392523 | 0.05902429 |
| Level Design | Hobbyist | 1.385000 | 0.04351255 |
| Marketing & PR | Professional | 1.514019 | 0.06818358 |
| Marketing & PR | Hobbyist | 1.505000 | 0.04962042 |
| Music | Professional | 1.345794 | 0.05491662 |
| Music | Hobbyist | 1.400000 | 0.04594075 |
| Programming | Professional | 1.925234 | 0.06989038 |
| Programming | Hobbyist | 1.960000 | 0.05391227 |
| Sound Design | Professional | 1.439252 | 0.06110943 |
| Sound Design | Hobbyist | 1.395000 | 0.04642901 |
| Storytelling | Professional | 1.523364 | 0.06685334 |
| Storytelling | Hobbyist | 1.575000 | 0.04778380 |
| Voice Overs & Acting | Professional | 1.663551 | 0.07144122 |
| Voice Overs & Acting | Hobbyist | 1.670000 | 0.05496001 |
Perceived Usefulness of Generative AI in Different Areas by Usage¶
In [ ]:
# Reset to source data
data <- read_excel(file_path)
data <- data %>%
mutate(across(
all_of(usefulness_columns),
~ usefulness_mapping[as.character(.)]
))
# Ensure the usage of Generative AI column is correctly labeled
data <- data %>%
mutate(UsesGenAI = factor(
data[[23]],
levels = c("Yes", "No")
))
# Remove rows with all NA values in usefulness columns
data <- data %>%
filter(rowSums(!is.na(select(., all_of(usefulness_columns)))) > 0)
# Pivot the data to a long format for ggplot2
usefulness_long <- data %>%
pivot_longer(
cols = all_of(usefulness_columns),
names_to = "Area",
values_to = "UsefulnessScore"
)
# Remove rows with NA values in UsefulnessScore or UsesGenAI
usefulness_long <- usefulness_long %>%
filter(!is.na(UsefulnessScore), !is.na(UsesGenAI))
# Calculate mean and standard error for each group
usefulness_summary <- usefulness_long %>%
group_by(Area, UsesGenAI) %>%
summarise(
MeanScore = mean(UsefulnessScore, na.rm = TRUE),
SE = sd(UsefulnessScore, na.rm = TRUE) / sqrt(n()),
.groups = "drop"
)
# Bar plot with error bars side by side
ggplot(usefulness_summary, aes(
x = Area,
y = MeanScore,
fill = UsesGenAI
)) +
geom_bar(
stat = "identity",
position = position_dodge(width = 0.9)
) +
geom_errorbar(
aes(
ymin = MeanScore - SE,
ymax = MeanScore + SE
),
width = 0.2,
position = position_dodge(width = 0.9)
) +
labs(
title = "Perceived Usefulness of Generative AI
in Different Areas by Usage",
x = "Area",
y = "Mean Usefulness Score
1 = Not useful
2 = Somewhat useful
3 = Very useful"
) +
theme(axis.text.x = element_text(angle = 45, hjust = 1))
# Calculate mean and standard error for each group
usefulness_summary <- usefulness_long %>%
group_by(Area, UsesGenAI) %>%
summarise(
MeanScore = mean(UsefulnessScore, na.rm = TRUE),
SE = sd(UsefulnessScore, na.rm = TRUE) / sqrt(n()),
.groups = "drop"
)
# Print the summarized data
usefulness_summary
| Area | UsesGenAI | MeanScore | SE |
|---|---|---|---|
| <chr> | <fct> | <dbl> | <dbl> |
| Art & Assets | Yes | 2.313559 | 0.06334529 |
| Art & Assets | No | 1.650794 | 0.04773149 |
| Community management | Yes | 1.398305 | 0.06034791 |
| Community management | No | 1.232804 | 0.03742102 |
| Game Design | Yes | 1.576271 | 0.06184987 |
| Game Design | No | 1.121693 | 0.02609801 |
| Initial prototyping | Yes | 2.296610 | 0.06734694 |
| Initial prototyping | No | 1.687831 | 0.05213607 |
| Level Design | Yes | 1.627119 | 0.06345663 |
| Level Design | No | 1.238095 | 0.03686375 |
| Marketing & PR | Yes | 1.745763 | 0.07037331 |
| Marketing & PR | No | 1.359788 | 0.04486913 |
| Music | Yes | 1.601695 | 0.06716441 |
| Music | No | 1.243386 | 0.03629382 |
| Programming | Yes | 2.372881 | 0.06113112 |
| Programming | No | 1.682540 | 0.04889652 |
| Sound Design | Yes | 1.610169 | 0.06917688 |
| Sound Design | No | 1.285714 | 0.03918978 |
| Storytelling | Yes | 1.906780 | 0.06480147 |
| Storytelling | No | 1.338624 | 0.04120499 |
| Voice Overs & Acting | Yes | 1.966102 | 0.07314484 |
| Voice Overs & Acting | No | 1.481481 | 0.04953493 |
Perceived Usefulness of Generative AI in Different Areas by Internal Stigma Perception¶
In [ ]:
# Reset to source data
data <- read_excel(file_path)
data <- data %>%
mutate(across(
all_of(usefulness_columns),
~ usefulness_mapping[as.character(.)]
))
# Ensure the stigma column is correctly labeled
data <- data %>%
mutate(StigmaGenAI = factor(
data[[33]],
levels = c("Yes", "No")
))
# Remove rows with all NA values in usefulness columns
data <- data %>%
filter(rowSums(!is.na(select(., all_of(usefulness_columns)))) > 0)
# Pivot the data to a long format for ggplot2
usefulness_long <- data %>%
pivot_longer(
cols = all_of(usefulness_columns),
names_to = "Area",
values_to = "UsefulnessScore"
)
# Remove rows with NA values in UsefulnessScore or StigmaGenAI
usefulness_long <- usefulness_long %>%
filter(!is.na(UsefulnessScore), !is.na(StigmaGenAI))
# Calculate mean and standard error for each group
usefulness_summary <- usefulness_long %>%
group_by(Area, StigmaGenAI) %>%
summarise(
MeanScore = mean(UsefulnessScore, na.rm = TRUE),
SE = sd(UsefulnessScore, na.rm = TRUE) / sqrt(n()),
.groups = "drop"
)
# Bar plot with error bars side by side
ggplot(usefulness_summary, aes(
x = Area,
y = MeanScore,
fill = StigmaGenAI
)) +
geom_bar(
stat = "identity",
position = position_dodge(width = 0.9)
) +
geom_errorbar(
aes(
ymin = MeanScore - SE,
ymax = MeanScore + SE
),
width = 0.2,
position = position_dodge(width = 0.9)
) +
labs(
title = "Perceived Usefulness of Generative AI
in Different Areas by Internal Stigma Perception",
x = "Area",
y = "Mean Usefulness Score
1 = Not useful
2 = Somewhat useful
3 = Very useful"
) +
theme(axis.text.x = element_text(angle = 45, hjust = 1))
# Print the usefulness summary to get results in text form
usefulness_summary
| Area | StigmaGenAI | MeanScore | SE |
|---|---|---|---|
| <chr> | <fct> | <dbl> | <dbl> |
| Art & Assets | Yes | 1.659420 | 0.06213554 |
| Art & Assets | No | 2.106509 | 0.05310944 |
| Community management | Yes | 1.224638 | 0.04365861 |
| Community management | No | 1.355030 | 0.04773624 |
| Game Design | Yes | 1.217391 | 0.04080426 |
| Game Design | No | 1.360947 | 0.04557878 |
| Initial prototyping | Yes | 1.724638 | 0.06404018 |
| Initial prototyping | No | 2.082840 | 0.05900229 |
| Level Design | Yes | 1.282609 | 0.04712077 |
| Level Design | No | 1.473373 | 0.04969978 |
| Marketing & PR | Yes | 1.405797 | 0.05504072 |
| Marketing & PR | No | 1.591716 | 0.05656045 |
| Music | Yes | 1.326087 | 0.04950089 |
| Music | No | 1.426036 | 0.05012152 |
| Programming | Yes | 1.826087 | 0.06333000 |
| Programming | No | 2.047337 | 0.05680679 |
| Sound Design | Yes | 1.347826 | 0.05209275 |
| Sound Design | No | 1.461538 | 0.05173790 |
| Storytelling | Yes | 1.398551 | 0.05298640 |
| Storytelling | No | 1.686391 | 0.05386535 |
| Voice Overs & Acting | Yes | 1.630435 | 0.06511152 |
| Voice Overs & Acting | No | 1.698225 | 0.05863379 |
Perceived Usefulness of Generative AI in Different Areas by External Stigma Perception¶
In [ ]:
# Reset to source data
data <- read_excel(file_path)
data <- data %>%
mutate(across(
all_of(usefulness_columns),
~ usefulness_mapping[as.character(.)]
))
# Ensure the external stigma column is correctly labeled
data <- data %>%
mutate(ExternalStigmaGenAI = factor(
data[[34]],
levels = c("Yes", "No")
))
# Remove rows with all NA values in usefulness columns
data <- data %>%
filter(rowSums(!is.na(select(., all_of(usefulness_columns)))) > 0)
# Pivot the data to a long format for ggplot2
usefulness_long <- data %>%
pivot_longer(
cols = all_of(usefulness_columns),
names_to = "Area",
values_to = "UsefulnessScore"
)
# Remove rows with NA values in UsefulnessScore or ExternalStigmaGenAI
usefulness_long <- usefulness_long %>%
filter(!is.na(UsefulnessScore), !is.na(ExternalStigmaGenAI))
# Calculate mean and standard error for each group
usefulness_summary <- usefulness_long %>%
group_by(Area, ExternalStigmaGenAI) %>%
summarise(
MeanScore = mean(UsefulnessScore, na.rm = TRUE),
SE = sd(UsefulnessScore, na.rm = TRUE) / sqrt(n()),
.groups = "drop"
)
# Bar plot with error bars side by side
ggplot(
usefulness_summary,
aes(
x = Area,
y = MeanScore,
fill = ExternalStigmaGenAI
)
) +
geom_bar(
stat = "identity",
position = position_dodge(width = 0.9)
) +
geom_errorbar(
aes(
ymin = MeanScore - SE,
ymax = MeanScore + SE
),
width = 0.2,
position = position_dodge(width = 0.9)
) +
labs(
title = "Perceived Usefulness of Generative AI
in Different Areas by External Stigma Perception",
x = "Area",
y = "Mean Usefulness Score
1 = Not useful
2 = Somewhat useful
3 = Very useful"
) +
theme(axis.text.x = element_text(angle = 45, hjust = 1))
# Print the usefulness summary to get results in text form
usefulness_summary
| Area | ExternalStigmaGenAI | MeanScore | SE |
|---|---|---|---|
| <chr> | <fct> | <dbl> | <dbl> |
| Art & Assets | Yes | 1.886555 | 0.04762504 |
| Art & Assets | No | 1.971014 | 0.09226480 |
| Community management | Yes | 1.256303 | 0.03506786 |
| Community management | No | 1.434783 | 0.08122638 |
| Game Design | Yes | 1.239496 | 0.03243718 |
| Game Design | No | 1.492754 | 0.07895089 |
| Initial prototyping | Yes | 1.869748 | 0.04963473 |
| Initial prototyping | No | 2.101449 | 0.09715678 |
| Level Design | Yes | 1.352941 | 0.03679173 |
| Level Design | No | 1.507246 | 0.08909699 |
| Marketing & PR | Yes | 1.470588 | 0.04320554 |
| Marketing & PR | No | 1.637681 | 0.09677453 |
| Music | Yes | 1.386555 | 0.04004150 |
| Music | No | 1.362319 | 0.07717026 |
| Programming | Yes | 1.932773 | 0.04743620 |
| Programming | No | 2.000000 | 0.09683834 |
| Sound Design | Yes | 1.411765 | 0.04117875 |
| Sound Design | No | 1.405797 | 0.08340280 |
| Storytelling | Yes | 1.512605 | 0.04282804 |
| Storytelling | No | 1.710145 | 0.08784007 |
| Voice Overs & Acting | Yes | 1.684874 | 0.05006211 |
| Voice Overs & Acting | No | 1.608696 | 0.08812093 |
Some questions about how your use of Generative AI impact different areas of your work?¶
Impact of Generative AI on Different Areas of Work¶
In [ ]:
# Reset to source data
data <- read_excel(file_path)
# Define impact columns
impact_columns <- c("Efficiency?", "Quality?", "Enjoyment?", "Creativity?")
# Convert impact columns to factors with specific levels
data <- data %>%
mutate(across(all_of(impact_columns), ~ factor(., levels = c(
"Much lower", "Lower", "Neutral", "Higher", "Much Higher"
))))
# Map categorical responses to numeric values
impact_mapping <- c(
"Much lower" = 1,
"Lower" = 2,
"Neutral" = 3,
"Higher" = 4,
"Much Higher" = 5
)
data <- data %>%
mutate(across(all_of(impact_columns), ~ impact_mapping[as.character(.)]))
# Calculate mean impact scores for each area
mean_impact_scores <- data %>%
summarise(across(all_of(impact_columns), ~ mean(.x, na.rm = TRUE))) %>%
pivot_longer(
cols = everything(),
names_to = "ImpactArea",
values_to = "MeanScore"
)
# Print mean impact scores to check for anomalies
print(mean_impact_scores)
# Bar plot of the mean impact scores
ggplot(
mean_impact_scores,
aes(x = ImpactArea, y = MeanScore, fill = ImpactArea)
) +
geom_bar(stat = "identity") +
labs(
title = "Mean Impact of Generative AI
on Different Areas of Work",
x = "Impact Area",
y = "Mean Impact Score
1 = Much lower
5 = Much Higher"
) +
scale_y_continuous(limits = c(0, 5), breaks = 1:5)
# Here's the boxplot
# Pivot the data to a long format for ggplot2
impact_long <- data %>%
pivot_longer(
cols = all_of(impact_columns),
names_to = "ImpactArea",
values_to = "ImpactScore"
)
# Remove rows with NA values in ImpactScore
impact_long <- impact_long %>%
filter(!is.na(ImpactScore))
# Box plot of the impact of Generative AI on different areas
ggplot(impact_long, aes(x = ImpactArea, y = ImpactScore, fill = ImpactArea)) +
geom_boxplot() +
labs(
title = "Impact of Generative AI
on Different Areas of Work",
x = "Impact Area",
y = "Impact Score
1 = Much lower
5 = Much Higher"
) +
scale_y_continuous(limits = c(1, 5), breaks = 1:5)
# A tibble: 4 x 2 ImpactArea MeanScore <chr> <dbl> 1 Efficiency? 3.97 2 Quality? 3.08 3 Enjoyment? 3.46 4 Creativity? 3.33
Impact of Generative AI on Different Areas by Lead Role¶
In [ ]:
# Reset to source data
data <- read_excel(file_path)
data <- data %>%
mutate(across(
all_of(impact_columns),
~ impact_mapping[as.character(.)]
))
# Pivot the data to a long format for ggplot2
impact_long <- data %>%
pivot_longer(
cols = all_of(impact_columns),
names_to = "ImpactArea",
values_to = "ImpactScore"
)
# Remove rows with NA values in ImpactScore or LeadRole
impact_long <- impact_long %>%
filter(!is.na(ImpactScore), !is.na(`Are you in a lead role?`))
# Calculate mean impact scores for each area and role
mean_impact_scores_by_role <- impact_long %>%
group_by(ImpactArea, `Are you in a lead role?`) %>%
summarise(
MeanScore = mean(ImpactScore, na.rm = TRUE),
SE = sd(ImpactScore, na.rm = TRUE) / sqrt(n()),
.groups = "drop"
)
# Rename the role column for better readability
colnames(mean_impact_scores_by_role)[2] <- "LeadRole"
# Print mean impact scores by role to check for anomalies
print(mean_impact_scores_by_role)
# Bar plot with error bars side by side
barplot <- ggplot(
mean_impact_scores_by_role,
aes(
x = ImpactArea,
y = MeanScore,
fill = LeadRole
)
) +
geom_bar(
stat = "identity",
position = position_dodge(width = 0.9)
) +
geom_errorbar(
aes(
ymin = MeanScore - SE,
ymax = MeanScore + SE
),
width = 0.2,
position = position_dodge(width = 0.9)
) +
labs(
title = "Mean Impact of Generative AI
on Different Areas by Lead Role",
x = "Impact Area",
y = "Mean Impact Score
1 = Much lower
5 = Much Higher"
)
# Box plot of the impact of Generative AI on different areas by lead role
boxplot <- ggplot(
impact_long,
aes(
x = ImpactArea,
y = ImpactScore,
fill = `Are you in a lead role?`
)
) +
geom_boxplot(
position = position_dodge(width = 0.9)
) +
labs(
title = "Impact of Generative AI
on Different Areas by Lead Role",
x = "Impact Area",
y = "Impact Score
1 = Much lower
5 = Much Higher"
)
# Print the plots
print(barplot)
print(boxplot)
# A tibble: 8 x 4 ImpactArea LeadRole MeanScore SE <chr> <chr> <dbl> <dbl> 1 Creativity? No 3.39 0.244 2 Creativity? Yes 3.04 0.236 3 Efficiency? No 4.06 0.262 4 Efficiency? Yes 3.96 0.189 5 Enjoyment? No 3 0.313 6 Enjoyment? Yes 3.26 0.211 7 Quality? No 3.17 0.305 8 Quality? Yes 2.74 0.189
