Graphical conventions: learning from iterated visual communication

How do people agree on ways of communicating visual concepts?

Note: This repo reflects the state of the codebase as of CogSci 2019. To access the codebase used in the subsequent journal article, please refer to: https://github.com/hawkrobe/graphical_conventions.

workflow

• Run human experiments

  • communication task (/experiments/refgame/draw_chairs/)
    • Input: Shapenet chair collection and experimental design
    • Output: Human sketches and viewer decisions over time, communication efficiency timecourse
  • recognition task (/experiments/recog/)
    • Input: Sketches from communication task and 3D objects
    • Output: Sketch recognizability in context (4 objects) w/o interaction history

• Analyze human task performance data

  • /analysis/ipynb/golden/generate_refgame_dataframe.py
    • Input: raw mongo database records
    • Output: tidy formatted dataframes containing key behavioral variables (XX.csv, BIS.csv)
  • /analysis/rmd/analyze_refgame_dataframe.Rmd
    • Input: tidy dataframe generated by generate_refgame_dataframe.py
    • Output: timecourse visualizations of key behavioral variables; results of linear mixed effects modeling of timecourse
  • /analysis/ipynb/golden/generate_recog_dataframe.py
    • Input: raw mongo database records
    • Output: tidy formatted dataframes containing key behavioral variables (XX.csv, BIS.csv)
  • /analysis/ipynb/golden/analyze_recog_dataframe.py
    • Input: tidy dataframe generated by generate_recog_dataframe.py
    • Output: timecourse visualizations of key behavioral variables; results of linear mixed effects modeling of timecourse

• Model-based analyses of sketch data

  • /analysis/golden/analyze_sketch_features.py
    • Input: sketch features generated by extract_sketch_features.py
    • Output: timecourse visualizations of key sketch feature variables

• Write paper

  • CogSci '19 paper