interactive-task-map-group-advantage.Rmd

---
output: html_document
runtime: shiny
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)

library(ggplot2)
library(plotly)
library(dplyr)
library(tidyr)
library(readr)
library(shiny)
library(rlang)
```

```{r load-data, echo=FALSE, message=FALSE, warning=FALSE}
# Load PCA base
task_map <- read_csv('./outputs/processed_data/task_map.csv')
# Load group advantage at condition level (has task, complexity, playerCount, strong, weak)
ga_cond <- read_csv('./outputs/processed_data/condition_level_group_advantage.csv')

# PCA of task map
pca <- prcomp(task_map[, -1], scale = TRUE)
pca_df <- as.data.frame(pca$x[, 1:2]) %>% mutate(task_pca = task_map$task)

# Normalize task names between sources for robust joining
norm_task <- function(x) {
  x <- tolower(x)
  x <- gsub('[^a-z0-9]+', ' ', x)
  trimws(x)
}

pca_df <- pca_df %>% mutate(task_norm = norm_task(.data$task_pca))

# Map GA task names to task_map names
name_map <- c(
  'Advertisement Writing'        = 'Advertisement Writing',
  'Allocating Resources'         = 'Allocating resources to programs',
  'Divergent Association'        = 'Divergent Association Task',
  'Guess the Correlation'        = 'Guessing the correlation',
  'Logic Problem'                = 'Logic Problem',
  'Moral Reasoning'              = 'Moral Reasoning (Disciplinary Action Case)',
  'Putting Food Into Categories' = 'Putting food into categories',
  'Random Dot Motion'            = 'Random dot motion',
  'Recall Association'           = 'Recall association',
  'Recall Word Lists'            = 'Recall word lists',
  'Room Assignment'              = 'Room assignment task',
  'Sudoku'                       = 'Sudoku',
  'Typing'                       = 'Typing game',
  'Unscramble Words'             = 'Unscramble words (anagrams)',
  'Whac a Mole'                  = 'Whac-A-Mole',
  'WildCam'                      = 'Wildcam Gorongosa (Zooniverse)',
  'Wildcat Wells'                = 'Wildcat Wells',
  'Wolf Goat Cabbage'            = 'Wolf, goat and cabbage transfer',
  'Word Construction'            = 'Word construction from a subset of letters',
  'Writing Story'                = 'Writing story'
)

ga_cond <- ga_cond %>% mutate(
  task_mapped = dplyr::recode(.data$task, !!!name_map, .default = .data$task),
  task_norm = norm_task(.data$task_mapped)
)

# Ensure complexity is ordered factor
ga_cond <- ga_cond %>% mutate(
  complexity = factor(.data$complexity, levels = c('Low', 'Medium', 'High'), ordered = TRUE),
  playerCount = as.factor(.data$playerCount)
)

# For default view, compute task-level means (across complexity and group sizes)
ga_task_means <- ga_cond %>%
  group_by(.data$task_norm) %>%
  summarise(
    task = dplyr::first(.data$task_mapped),
    strong = mean(.data$strong, na.rm = TRUE),
    weak = mean(.data$weak, na.rm = TRUE),
    .groups = 'drop'
  )

# Join PCA with GA info; keep the mapped task name
pca_ga <- pca_df %>%
  inner_join(ga_task_means, by = 'task_norm') %>%
  transmute(PC1 = .data$PC1, PC2 = .data$PC2, task_norm = .data$task_norm,
            task = .data$task, strong = .data$strong, weak = .data$weak)

# Keep a version with all condition rows for filtering
pca_ga_cond <- pca_df %>% inner_join(ga_cond, by = 'task_norm')
```

```{r ui-server, echo=FALSE}
ui <- fluidPage(
  titlePanel('Interactive Task Map — Group Advantage (20 tasks)'),
  sidebarLayout(
    sidebarPanel(
      radioButtons('dv', 'Color by:', choices = c('Strong' = 'strong', 'Weak' = 'weak'), selected = 'strong', inline = TRUE),
  selectInput('complexity', 'Complexity:', choices = c('All', 'Low', 'Medium', 'High'), selected = 'All'),
  checkboxGroupInput('groupSize', 'Group size:', choices = c('3', '6'), selected = c('3','6'), inline = TRUE),
  checkboxInput('showLabels', 'Show labels', value = FALSE),
  helpText('Note: Colors are centered at 1. Blue indicates advantage (>1), red indicates disadvantage (<1).')
    ),
    mainPanel(
      plotlyOutput('map_plot')
    )
  )
)

server <- function(input, output) {
  # reactive filtered data
  filtered_points <- reactive({
    dv_col <- input$dv

    if (input$complexity == 'All') {
      df <- pca_ga %>% mutate(value = .data[[dv_col]])
    } else {
      df <- pca_ga_cond %>%
        filter(.data$complexity == input$complexity) %>%
        filter(as.character(.data$playerCount) %in% input$groupSize) %>%
        group_by(.data$task_norm, .data$task_mapped, .data$PC1, .data$PC2) %>%
        summarise(value = mean(.data[[dv_col]], na.rm = TRUE), .groups = 'drop') %>%
        rename(task = .data$task_mapped)
    }

    df
  })

  output$map_plot <- renderPlotly({
    dv_label <- ifelse(input$dv == 'strong', 'Strong Advantage', 'Weak Advantage')

    base <- ggplot() +
      geom_point(data = pca_df, aes(x = .data$PC1, y = .data$PC2), color = 'grey70', alpha = 0.5, size = 2) +
      theme_minimal() + labs(x = 'PC1', y = 'PC2')

    pts <- filtered_points()

    g <- base +
      geom_point(data = pts, aes(x = .data$PC1, y = .data$PC2, color = .data$value, text = paste0(.data$task, '\n', dv_label, ': ', round(.data$value, 3))), size = 4) +
      scale_color_gradient2(name = dv_label, low = '#b2182b', mid = '#f7f7f7', high = '#2166ac', midpoint = 1)

    if (isTRUE(input$showLabels)) {
      g <- g + geom_text(data = pts, aes(x = .data$PC1, y = .data$PC2, label = .data$task), vjust = -0.8, size = 3)
    }

    ggplotly(g, tooltip = c('text'))
  })
}

shinyApp(ui, server)
```