Simulating Attention and Performance tasks with an LLM

Author

Affiliation

Matthew J. C. Crump

Brooklyn College of CUNY

Abstract

A computational modeling project examining whether or not a GPT model can simulate human performance in classic attention and performance tasks.

Project Information

The purpose of this project is to assess capabilities of LLMs, such as OpenAI’s GPT models, to simulate behavioral data from classic attention and performance tasks in cognitive psychology, like the Stroop task.

Status:

• In progress
• Exploratory

Repository Information

This is a website for a cognitive psychology research project. The aim of this website is to create and share reproducible research assets throughout the project lifespan. This repository is built with quarto. Source code is available from the github repository for this project.

Dependencies:

List of R packages used.

# get packages and print version in tibble
dependencies <- renv::dependencies()

Finding R package dependencies ... Done!

unique_dependencies <- unique(dependencies$Package)

dependency_tibble <- sapply(unique_dependencies,
       FUN = function(x) as.character(packageVersion(x))) |>
  tibble::enframe()

knitr::kable(dependency_tibble)

name	value
knitr	1.42
renv	0.17.3
tibble	3.1.8
rmarkdown	2.20
openai	0.4.1
patchwork	1.1.2
tidyverse	1.3.2
xtable	1.8.4
stringr	1.5.0
papaja	0.1.1