Multi-party tangrams

Reproducible paper pipeline

write-ups/manuscript/tangrams-manuscript.Rmd and write-ups/manuscript/supplement.Rmd should render. Due to emojis in tables, you will need to use LuaLaTex to go tex -> pdf which may be easiest in an external tex editor.

What gets used:

• manuscripts sources code/prep_ms.R which reads processed data files and combines
• model summaries and predictions are saved in code/paper_mods

Fuller pipeline:

• prep_processed_data.R - goes from raw empirica files to processed data files (note that it expects hand filtering of chat to occur at various points)
• run_sbert.R, sbert_play.ipynb, use_sbert_embeddings.R - prep for sbert embedding, embed, take sbert embeddings and return processed cosine sims
• mods_for_paper.R, additional_models.R, mega-analytic_mods.R -- model code for all models
• preds_for_paper.R takes models and outputs summarized predictions for plots
• prep_ms.R reads lots of processed data and does joins -- sourced by ms and supplement

Guide to repo organization

• code: R and Py code for analyses
• models contains large model files
• data: data files (un & pre-processed): study 1, 2a/2b/2c, 3 are relevant subfolders
• experiments: code and stuff to run experiments
• write-ups: paper-like things

Scripts and write-ups not described above may be broken / break things b/c of refactoring that has happened. Their use is deprecated.

Testing experiments

To develop locally,

1. make sure you have meteor installed,
2. clone the repo, and run meteor npm install to get the dependencies
3. launch locally with meteor --settings local.json (the default admin password is password -- change this in local.json.
4. go to http://localhost:3000/admin in your browser