Adaptive computation Analysis Repo

Collection of analysis scripts used in "Adaptive computation as a new mechanism of dynamic human attention"

@article{Belledonne_Butkus_Scholl_Yildirim,
title={Adaptive computation as a new mechanism of dynamic human attention},
author={Belledonne, Mario and Butkus, Eivinas and Scholl, Brian J. and Yildirim, Ilker},
journal={Psychological Review},
year={2025}
} 

Replication

The main reports used in the submission are located under reports/. These are organized by manuscript section. These reports use some preprocessing scripts (e.g., parsing human responses) located under scripts/. This also includes scripts to create some of the figures.

The human and model data can be downloaded as a tar.gz archive here. To extract the dataset run tar -xzf data.tar.gz. The result should be a folder called data/ with three sections corresponding to the manuscript layout (the probes, effort, and accuracy experiments).

Dependencies for main analysis

The main reports used in the submission depend on any recent version of R-studio, in particular the tidyverse and ksmooth packages.

A docker enviroment used for the submission is provided at Dockerfile under the project root.

Dependencies for parsing raw human behavior

The raw behavioral data (collected via https://github.com/CNCLgithub/mot-psiturk) can be parsed by scripts/parsed_db_exp_[probes|effort].py. These will require a python environment with the following dependencies:

numpy
pandas
sqlalchemy