EXPLORING STEREOTYPING OF SMOKERS RELATED TO AGE, SPORTS, AND ENVIRONMENTAL AWARENESS IN AUTOMATIC TEXT GENERATION PROGRAMS.
Keywords
Loading...
Authors
Issue Date
2024-03-13
Language
en
Document type
Journal Title
Journal ISSN
Volume Title
Publisher
Title
ISSN
Volume
Issue
Startpage
Endpage
DOI
Abstract
This study aimed to explore stereotyping in automatic text generation systems. The main question it intended to answer was: ‘To what extent does the mention of a smoking habit in a prompt given to an automatic text generating system like ChatGPT have an effect on the stereotypical language generated in the output?’ This research contributes to the field by gathering as much data as possible for future projects to be able to account for possible stereotyping in the program ChatGPT. To answer the research question specific prompts were constructed and a codebook was made with rules according to which the final dataset was annotated. For stereotyping this study focussed on smoking habit, age and stereotypical utterances concerning the concepts “sportive” and “environmentally aware”. This study found that ChatGPT-3.5 considers the persona descriptions it generated that were smokers to be less concerned with the environment and sports, moreover these personas were on average older than the personas it generated that were nonsmokers.
Description
Citation
Supervisor
Faculty
Faculteit der Letteren