--- title: "Base Study 2 — Quantitative Analysis" author: "Gian Luca Liehner" date: "`r format(Sys.Date(), '%B %d, %Y')`" format: html: toc: true toc-location: left editor: source --- ```{r setup} #| message: false #| echo: false knitr::opts_chunk$set(cache = TRUE, message = TRUE) # Loading packages library(tidyverse) # Data manipulation library(questionr) # Frequency tables library(kableExtra) # Nice kable tables library(janitor) # Make databases look nice library(careless) # Longstrings library(psych) # Scale manipulation library(Hmisc) # Social data manipulation library(corrplot) # Correlation plots library(likert) # Visualising likert scales library(lubridate) # Date manipulation library(likert) library(sjPlot) library(sjmisc) # Theme used for dissertation theme_dissertation <- function () { theme_minimal(base_size=12) + theme( panel.background = element_blank(), legend.background = element_rect(fill="transparent", colour=NA), legend.key = element_rect(fill="transparent", colour=NA), text=element_text(size=12, family="Gill Sans"), plot.background = element_rect(fill = "transparent", color = NA), axis.title.y = element_text(margin = margin(t = 0, r = 20, b = 0, l = 0), size = 12), axis.title.x = element_text(margin = margin(t = 10, r = 0, b = 0, l = 0), size = 12), plot.title = element_text(margin=margin(0,0,10,0)), legend.margin=margin(l = 10,), #panel.grid.major = element_line(colour = "#565555", size = .25) plot.caption = element_text(size=8, color = "#333333", face = "italic"), plot.subtitle = element_text(size=8, color = "#333333") ) } # Colours colour_dis <- c( teal = "#006d77", mint = "#83c5be", off_white = "#edf6f9", peach = "#ffddd2", clay = "#e29578", blue = "#457b9d", blue_light = "#a8dadc", pink = "#ef476f", pink_light = "#ffb4c8" ) ``` # Loading data ```{r loading-data} #| warning: false raw_data <- read_csv("raw_data.csv") n = raw_data %>% tally() ``` # Overview of raw data Data before cleaning and with a the data of today (2025/10/13) has N = `r raw_data %>% tally()`. ```{r overview} is.na(raw_data) %>% colSums() %>% kable(caption = "List of all variables before cleaning including NAs", col.names = c("Variable name")) %>% kable_styling(bootstrap_options = c("striped", "hover")) %>% scroll_box(height = "300px") ``` # Data Cleaning ## Screening Only accepting surveys that are matching the screening criteria ```{r screening} data <- raw_data %>% filter(quality_control == "complete") %>% filter(Progress == 100) %>% filter(Finished == 1) %>% filter(is.na(Q_TerminateFlag)) %>% filter(StartDate >= lubridate::as_datetime("2022-04-28 18:00:00")) ``` Dropped cases `r data %>% tally()-n` `r n <- data %>% tally()` ## Removing duplicated respondents ```{r remove-double-replies} # If respondent ID is duplicated in the data, remove entries data <- data %>% arrange(rid) %>% # optional: ensures deterministic order distinct(rid, .keep_all = TRUE) ``` Dropped cases `r data %>% tally()-n` `r n <- data %>% tally()` ## Filtering out minors ```{r minors} data <- data %>% filter(age >= 18) ``` Dropped cases `r data %>% tally()-n` `r n <- data %>% tally()` ## Removing speeders ```{r median-time} #| echo: false median_completion_time <- median(as.numeric(data$`Duration (in seconds)`)) ``` Median survey time is `r median_completion_time` seconds (`r round(median_completion_time/60,2)` minutes) ```{r speeders} data <- data %>% filter(!as.numeric(`Duration (in seconds)`) < median_completion_time/2) ``` Dropped cases `r data %>% tally()-n` `r n <- data %>% tally()` # Data manipulation ## Collapsing experiment conditions ```{r collapsing} #| include: false # Extracting condition data <- data %>% mutate(condition = case_when( watched_video_contro == "1" ~ "control", watched_video_glo == "1" ~ "global", watched_video_local == "1" ~ "local", watched_video_mixed == "1" ~ "mixed", TRUE ~ NA_character_ )) # Collapsing use intention data <- data %>% mutate(across(starts_with("use_intention_"), as.numeric)) %>% mutate( use_intention_1 = coalesce(!!!select(., starts_with("use_intention_1"))), use_intention_2 = coalesce(!!!select(., starts_with("use_intention_2"))), use_intention_3 = coalesce(!!!select(., starts_with("use_intention_3"))), use_intention_4 = coalesce(!!!select(., starts_with("use_intention_4"))), use_intention_5 = coalesce(!!!select(., starts_with("use_intention_5"))) ) # Collapsing trust data <- data %>% mutate(across(starts_with("trust_"), as.numeric)) %>% mutate( trust_1 = coalesce(!!!select(., starts_with("trust_1"))), trust_2 = coalesce(!!!select(., starts_with("trust_2"))), trust_3 = coalesce(!!!select(., starts_with("trust_3"))), trust_4 = coalesce(!!!select(., starts_with("trust_4"))), trust_5 = coalesce(!!!select(., starts_with("trust_5"))), trust_6 = coalesce(!!!select(., starts_with("trust_6"))), trust_7 = coalesce(!!!select(., starts_with("trust_7"))), trust_8 = coalesce(!!!select(., starts_with("trust_8"))), trust_9 = coalesce(!!!select(., starts_with("trust_9"))), ) # Collapsing taigers data <- data %>% mutate(across(starts_with("taigers_"), as.numeric)) %>% mutate( taigers_1 = coalesce(!!!select(., starts_with("taigers_1"))), taigers_2 = coalesce(!!!select(., starts_with("taigers_2"))), taigers_3 = coalesce(!!!select(., starts_with("taigers_3"))), taigers_4 = coalesce(!!!select(., starts_with("taigers_4"))) ) ``` New variable names are: - `condition` for the experimental condition - `use_intention_1` to `use_intention_5` - `trust_1` to `trust_9` - `taigers_1` to `taigers_4` ## Data selection Selecting only relevant data for dissertation ```{r data-selection} data <- data %>% select( ResponseId, age, gender, school_edu, professional_edu, professional_edu_text = professional_edu_4_TEXT, jobtype, starts_with("dispo_trust"), starts_with("locus_control"), starts_with("ati"), starts_with("mental_model"), starts_with("ai_exp"), starts_with("gaais"), condition, use_intention_1:use_intention_5, trust_1:trust_9, taigers_1:taigers_4, rid ) ``` ## Longstring cleaning ```{r longstring} # Locus of control data <- data %>% filter(longstring(select(., starts_with("locus"))) != 4) data <- data %>% filter(longstring(select(., starts_with("ati"))) != 9) data <- data %>% filter(longstring(select(., starts_with("gaais"))) != 20) ``` Dropped cases `r data %>% tally()-n` `r n <- data %>% tally()` --- **Final N:** `r n` ## Recoding data ```{r conversion} # Converting to numeric data <- data %>% mutate(across(c(-ResponseId, -professional_edu_text, -condition), ~ as.numeric(.x))) # Recoding gender data$gender <- factor(data$gender, levels = c(1,2,3,4), labels = c("male", "female", "diverse", "not specified" ) ) data$school_edu <- factor(data$school_edu, levels = c(1,2,3,4), labels = c("No education degree", "Secondary school degree (i.e. Hauptschulabschluss)", "Secondary school degree (i.e. Realschulabschluss)", "High school diploma (e.g Abitur)") ) data$professional_edu <- factor(data$professional_edu, levels = c(1,2,3), labels = c("No professional education degree", "Completed apprenticeship", "Completed studies") ) data$jobtype <- factor(data$jobtype, levels = c(1,2,3,4,5,6,7), labels = c("Student/ apprenticeship", "Employee & Student", "Employee", "Employer", "Self employed without employees", "Retired", "currently unemployed") ) ``` # Sample description ## Age ```{r age} #| echo: false psych::describe(data$age, fast = TRUE) %>% kable(caption = "Descriptives for age", digits = 2, row.names = FALSE) %>% kable_material() data %>% ggplot + aes(age) + geom_histogram(binwidth = 1, fill=colour_dis["mint"], color=colour_dis["teal"], size=.3) + geom_density(aes(y = ..density..*nrow(data)), color=colour_dis["pink"]) + geom_vline(xintercept = mean(data$age), linetype="dotted", color = colour_dis["pink"], size=1.5) + labs(x="Age", title = "Age histrogram", y="Absolute count", caption = "Blue dotted line represents mean, and pink solid line represents density") + theme_dissertation() ``` ## Gender ```{r gender} #| echo: false questionr::freq(data$gender, sort = "dec") %>% kable(caption = "Frequency table for gender", digits = 2) %>% kable_material() ``` ## School education ```{r school_edu} #| echo: false questionr::freq(data$school_edu, sort = "dec") %>% kable(caption = "Frequency table for school education", digits = 2) %>% kable_material() ``` ## Professional education ```{r prof_edu} #| echo: false questionr::freq(data$professional_edu, sort = "dec") %>% kable(caption = "Frequency table for professional education", digits = 2) %>% kable_material() ``` ```{r prof_edu_others} #| echo: false data %>% filter(!is.na(professional_edu_text)) %>% select(professional_edu_text) %>% kable(caption = "Other education levels for professional education") %>% kable_material() ``` ## Occupation ```{r occupation} #| echo: false questionr::freq(data$jobtype, sort = "dec") %>% kable(caption = "Frequency table for current occupation", digits = 2) %>% kable_material() ``` # Personality traits ## Locus of control ```{r control} #| echo: false control_items <- data %>% select(starts_with("locus")) control_keys <- list(control_scores = c( "locus_control_1", "locus_control_2", "-locus_control_3_n", "-locus_control_4_n" )) control_scores <- scoreItems(control_keys, control_items, min = 1, max = 6) data <- data %>% bind_cols(control_scores$scores %>% as_tibble()) control_scores$scores %>% psych::describe(fast = TRUE, ranges = FALSE) %>% kable(caption = "Scores for scale 'locus of control'", digits = 2, row.names = FALSE) %>% kable_material() ``` The reliability for the *locus of control scale* is `r round(control_scores$alpha, 2)`. ## Disposition to trust ```{r dispo-trust} #| echo: false dispo_trust_items <- data %>% select(starts_with("dispo_trust")) dispo_trust_keys <- list(dispo_trust_scores = c( "dispo_trust_1", "dispo_trust_2", "dispo_trust_3", "dispo_trust_4" )) dispo_trust_scores <- scoreItems(dispo_trust_keys, dispo_trust_items, min = 1, max = 6) data <- data %>% bind_cols(dispo_trust_scores$scores %>% as_tibble()) dispo_trust_scores$scores %>% psych::describe(fast = TRUE, ranges = FALSE) %>% kable(caption = "Scores for scale 'trust disposition'", digits = 2, row.names = FALSE) %>% kable_material() ``` The reliability for the *disposition to trust scale* is `r round(dispo_trust_scores$alpha, 2)` ## Affinity for technology ```{r ati} #| echo: false ati_items <- data %>% select(starts_with("ati")) ati_keys <- list(ati_scores = c( "ati_1", "ati_2", "-ati_3_n", "ati_4", "ati_5", "-ati_6_n", "ati_7", "-ati_8_n", "ati_9" )) ati_scores <- scoreItems(ati_keys, ati_items, min = 1, max = 6) data <- data %>% bind_cols(ati_scores$scores %>% as_tibble()) ati_scores$scores %>% psych::describe(fast = TRUE, ranges = FALSE) %>% kable(caption = "Scores for scale 'affinity for technology'", digits = 2, row.names = FALSE) %>% kable_material() ``` The reliability for the *affinty for technology scale* is `r round(ati_scores$alpha, 2)` ## Experience with AI ```{r exp} #| echo: false ai_exp_items <- data %>% select(starts_with("ai_exp")) ai_exp_keys <- list(ai_exp_scores = c( "ai_exp_1", "ai_exp_2", "ai_exp_3", "ai_exp_4", "ai_exp_5", "ai_exp_6", "ai_exp_7", "ai_exp_8", "ai_exp_9" )) ai_exp_scores <- scoreItems(ai_exp_keys, ai_exp_items, min = 1, max = 6) data <- data %>% bind_cols(ai_exp_scores$scores %>% as_tibble()) ai_exp_scores$scores %>% psych::describe(fast = TRUE, ranges = FALSE) %>% kable(caption = "Scores for scale 'experience with AI'", digits = 2, row.names = FALSE) %>% kable_material() ``` The reliability for the *Experience with AI scale* is `r round(ai_exp_scores$alpha, 2)` ## General attitude towards AI (GAAIS) ```{r gaais} #| echo: false gaais_items <- data %>% select(starts_with("gaais")) gaais_keys <- list(gaais_scores = c( "gaais_1", "gaais_2", "-gaais_3_n", "gaais_4", "gaais_5", "-gaais_6_n", "gaais_7", "gaais_8_n", "-gaais_9_n", "-gaais_10_n", "gaais_11", "gaais_12", "gaais_13", "gaais_14", "-gaais_15_n", "gaais_16", "gaais_17", "gaais_18", "-gaais_19_n", "-gaais_20_n" )) gaais_scores <- scoreItems(gaais_keys, gaais_items, min = 1, max = 6) data <- data %>% bind_cols(gaais_scores$scores %>% as_tibble()) gaais_scores$scores %>% psych::describe(fast = TRUE, ranges = FALSE) %>% kable(caption = "Scores for scale 'GAAIS'", digits = 2, row.names = FALSE) %>% kable_material() ``` The reliability for the *GAAIS scale* is `r round(gaais_scores$alpha, 2)` # Descriptives ## Mental models ```{r} #| echo: false #| message: false data %>% select(starts_with("mental")) -> mental_models_likert_data likert_plot_mental_models<- plot_likert(mental_models_likert_data, values = "sum.outside", show.prc.sign = TRUE, wrap.labels = 64, title = "Mental models of AI", reverse.scale = TRUE, axis.titles=rev(c("Level of agreement", "Statements about artificial intelligence")), axis.labels=c( "3. AI is a software tool like any other.", "2. AI is like a team member", "4. AI is nothing more than linear algebra.", "1. AI has something magical, \n superhuman about it." ), sort.frq = "pos.asc", show.n = FALSE) + theme_dissertation()+ scale_fill_manual(labels=c("strongly disagree", "disagree", "rather disagree", "rather agree", "agree", "strongly agree"), values=c("#ed6976", "#f18c90", "#f5acac", "#a6d5c1", "#81c6ab", "#50b796"))+ theme(legend.position="bottom", text = element_text(size = 20)) # Legende unten likert_plot_mental_models ggsave('./exports/mental_models_ai.png', width = 14, height = 9, units = "cm", scale = 2) # Speichern der Grafik als PNG ggsave('./exports/mental_models_ai.svg', width = 14, height = 9, units = "cm", scale = 2) # Speichern der Grafik als PNG ``` ## Trust ```{r trust} #| echo: false trust_items <- data %>% select(starts_with("trust")) trust_keys <- list(trust_scores = c( "-trust_1", "-trust_2", "trust_3", "trust_4", "trust_5", "trust_6", "trust_7", "-trust_8", "-trust_9" )) trust_scores <- scoreItems(trust_keys, trust_items, min = 1, max = 6) data <- data %>% bind_cols(trust_scores$scores %>% as_tibble()) trust_scores$scores %>% psych::describe(fast = TRUE, ranges = FALSE) %>% kable(caption = "Scores for scale 'trust in AI'", digits = 2, row.names = FALSE) %>% kable_material() ``` The reliability for the *Trust in AI* is `r round(trust_scores$alpha, 2)` ## Use intention ```{r use-intention} #| echo: false use_intention_items <- data %>% select(starts_with("use_intention")) use_intention_keys <- list(use_intention_scores = c( "use_intention_1", "-use_intention_2", "use_intention_3", "use_intention_4" )) use_intention_scores <- scoreItems(use_intention_keys, use_intention_items, min = 1, max = 6) data <- data %>% bind_cols(use_intention_scores$scores %>% as_tibble()) use_intention_scores$scores %>% psych::describe(fast = TRUE, ranges = FALSE) %>% kable(caption = "Scores for scale 'use intention of AI'", digits = 2, row.names = FALSE) %>% kable_material() ``` # Interference The reliability for the *Use intention of AI* is `r round(use_intention_scores$alpha, 2)` ```{r recoding-explanation} #| echo: false data <- data %>% mutate(explanation = if_else(condition == "control", 0, 1)) %>% mutate(global = if_else(condition == "global" | condition == "mixed", 1, 0)) %>% mutate(local = if_else(condition == "local" | condition == "mixed", 1, 0)) %>% mutate(explanation = factor(explanation, levels = c(0,1), labels = c("No explanation", "Explanation") ) ) ``` ## Model 1: Experiment condition ```{r} model_1 <- lm(cbind(trust_scores, use_intention_scores) ~ local * global, data) summary(model_1) ``` ## Model 2: User diversity ```{r model-2} model_2 <- lm(cbind(trust_scores, use_intention_scores) ~ local * global + age + dispo_trust_scores + ati_scores + gaais_scores, data) summary(model_2, test="Pillai") ``` ```{r} anova(model_1, model_2) summary(model_2)$r.squared ``` # Exkurs Also evaluating the TAIGERS variable ## TAIGERS variable ```{r taigers} #| echo: false taigers_items <- data %>% select(taigers_1, taigers_2, taigers_3, taigers_4) taigers_keys <- list(taigers_scores = c( "taigers_1", "taigers_2", "taigers_3", "taigers_4" )) taigers_scores <- scoreItems(taigers_keys, taigers_items, min = 1, max = 6) data <- data %>% bind_cols(taigers_scores$scores %>% as_tibble()) taigers_scores$scores %>% psych::describe(fast = TRUE, ranges = FALSE) %>% kable(caption = "Scores for scale 'understandability'", digits = 2, row.names = FALSE) %>% kable_material() ``` The reliability for the *explanatory power* is `r round(taigers_scores$alpha, 2)` ```{r} model_3 <- manova(cbind(trust_scores, use_intention_scores, taigers_scores) ~ local * global + age + dispo_trust_scores + ati_scores + gaais_scores, data) summary(model_3, test = "Pillai") summary.aov(model_2) anova(model_1, model_2) ```