This repository contains code and data from an online study that sought to measure some of the contours of virtual enfreakment in fighting games (Street Fighter V and Tekken 7 specifically). You can browse all of the files online at: https://github.com/solsword/enfreakment-study/
This document explains what resources are available; the details of the survey methodology and statistical analysis are explained in this methodology document.
The turk/ folder contains the HTML templates and other data required to construct the Amazon Mechanical Turk Human Intelligence Tasks (HITs) that we used to carry out the study. There is code there which was used to create randomized HITs and put together templates, etc. The Makefile in that directory, although not well-documented, should give a sense of how things fit together; it relies on the clingo constraint solving library to create counterbalanced character groupings, as well as python for some of the scripts.
If you want to see an example of a fully constructed HIT, the turk/template-local.html file shows what a single HIT template looks like when filled out; if you are viewing it locally you should run make template-local.html in the turk/ folder first otherwise the images will not show up.
The data/ folder contains both anonymized survey response data (in the file efr.tsv) as well as code for analyzing it based on our hypotheses. Again, the Makefile shows how things fit together; note that there are some rules there for processing the raw data, but based on our study protocols this raw data which contains AMT user IDs cannot be made public.
The analysis code uses the scipy and krippendorf Python packages, and should work with Python versions roughly 3.5-5.8 at least. It uses fairly large bootstrap samples to do statistics, and can take dozens of minutes or more to finish, although it does print progress messages (try tail -f report.txt while make report.txt is running).
The Makefile scripts assume a POSIX environment and probably a few GNU utilities on top of that, so they’re easiest to run in a Mac or Linux environment, but you could run some of the data processing steps manually if you wanted to, and if you’re designing a similar study or if you just want to run your own analysis of the data, you can likely simplify things enormously.