# Overview This OSF repository contains the data and materials for the project, "The Task Space: A Multidimensional Representation of Team Tasks." # Key Files - **24_dimensions_clean.csv**: This CSV lists the descriptions and sources of all 24 dimensions currently included in our "map" of tasks. - **102_tasks_with_sources_clean.csv**: This CSV lists the names, sources, and descriptions of all 102 included tasks. - **questions_displayed_to_raters.csv**: This CSV lists the question and elaboration text for rating tasks on each dimension. This information was displayed to our rating panel. - **raw_map.csv** This CSV contains the raw rating data. - **task_map.csv** This CSV is a cleaned version of the rated tasks, presented as a 102 (task) x 24 (dimension) matrix. It can be generated from the raw data using the script `analysis_scripts/generate_task_map_from_raw.Rmd`. - **task-map.png**: This is a simple PCA plot of our 102 tasks. We chose to visualize the current plot using k = 3 clusters, clustered using k-means. This image can be generated from the code included in `analysis_scripts/clean_task_map_visuals.Rmd`. # Key Folders - **rater_training/**: This folder contains the .qsf and .pdf of the pre-test and training materials that we used to develop our MTurk rating panel. - **analysis_scripts/**: This folder contains five scripts that replicate our data cleaning process and the process by which we generated the figures in the paper: - **v0_synthetic_data_figures/**: This folder contains some of the original analyses performed using the Task Space on synthetic data. As of August 2024, they are no longer included in the manuscript, but remain here for reference. - **v1_empirical_data_figures_analysis/**: This folder contains the necessary data and a Jupyter Notebook (.ipynb) to replicate the analysis figures of the empirical case study (newly added to the manuscript as of August 2024). # Key Scripts Code in this repository is written in a combination of R and Python, with some of the basic analyses (generating the Task Map from the raw ratings) done in R and some of the more advanced analyses (particularly modeling and the empirical case study) done in Python. 1. **generate_task_map_from_raw.Rmd** is the basic data cleaning file that takes the raw rating data and generates the task map; 2. **clean_task_map_visuals.Rmd** generates the task map image (`task-map.png`), as well as other visuals; 3. **rr_analyses_final.ipynb** contains the analysis code for the empirical case study and visuals that were added in the R&R processes (new as of August 2024). The following files pertain to an illustration using synthetic data that is no longer in the manuscript as of August 2024: 4. **choose-10-tasks.ipynb** is a Python file that uses Euclidean distance to select tasks that are as close together or as far away as possible (this scripts generates the tasks that are then featured in our figures). This Python file is standalone; its results are hard-coded into `clean_task_map_visuals.Rmd`, but it is here to demonstrate reproducibility. 5. **svm_with_synthetic_data.ipynb** implements the Support Vector Machine described in the paper, as part of our simulation of an experiment with 10 tasks. The simulated data is... 6. **tasks_with_synthetic_data.csv**, in this folder, and it can be generated by running `clean_task_map_visuals.Rmd`.