Supplementary materials for the article: Using ChatGPT for Human Computer Interaction Research: A Primer



Supplementary data for the paperTabone, W., & De Winter, J. C. F. (2023). Using ChatGPT for Human–Computer Interaction Research: A Primer. Royal Society Open Science, 10, 231053.

This repository contains MATLAB scripts (tested in MATLAB R2023a), input data (source data read by the MATLAB scripts), and saved ChatGPT outputs for Study 1, 2, and 3 of the following research paper:

- Study1.m contains code for 2 different prompts, as well as different batch sizes (25 vs. 992), as described in the paper.
- Study1_bootstrapping.m contains the code for the bootstrapping approach, also described in the paper.
- Study1_randomness_test.m contains the code corresponding to the systematic variation of the temperature parameter described in the paper.
- Study2.m contains the code that corresponds to the interview summary for the Virtual fence augmented reality (AR) interface.
- Study2_16k_test.m contains the code of a trial, in which the entire interview was submitted all at once; something that has recently become possible with the 16k variant of the GPT-3.5 (not described in the paper)
- Study2_content_analysis.m and Study2_content_analysis_16k contain code for attempts at letting GPT perform a content analysis of the interview (described in the discussion of the paper).
- Study2_gpt4.m is similar to Study2.m but now adopts GPT-4 instead of GPT-3.5
- Study3_bootstrapping.m is code for the bootstrapping approach of the transcripts, as described in the paper.
Date made available4 Sept 2023
PublisherTU Delft - 4TU.ResearchData

Cite this