國立虎尾科技大學 |

From Wiki or ChatGPT? Perceptions of AI-Generated Text /

Record Type:	Language materials, printed : Monograph/item
Title/Author:	From Wiki or ChatGPT? Perceptions of AI-Generated Text // Brianna Elizabeth O'Boyle.
Author:	O'Boyle, Brianna Elizabeth,
Description:	1 electronic resource (50 pages)
Notes:	Source: Masters Abstracts International, Volume: 85-10.
Contained By:	Masters Abstracts International85-10.
Subject:	Linguistics. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31238467
ISBN:	9798382305486

From Wiki or ChatGPT? Perceptions of AI-Generated Text /
O'Boyle, Brianna Elizabeth,

From Wiki or ChatGPT? Perceptions of AI-Generated Text /Brianna Elizabeth O'Boyle. - 1 electronic resource (50 pages)

Source: Masters Abstracts International, Volume: 85-10.

Arguably, one of the most revolutionary technological advancements of the 2020s has been the creation of ChatGPT, an AI-powered chatbot aided by innovations in the field of Natural Language Processing and the development of Large Language Models. There have been many people who use ChatGPT for its ability to generate text within seconds, but there remain questions of how do humans perceive AI-generated text, and are there certain genres of text people believe ChatGPT is better at generating. This study investigates human perception of AI-generated text using methodological frameworks from Sociolinguistics, specifically with a matched guise task. Participants were shown ten written texts of two genres, five expository texts and five creative texts, and asked to rate the texts based on three statements using a Likert scale. Participants were told the texts came from a variety of sources, but all of them were generated by ChatGPT. It was found in the majority of models participants were sensitive to the label of the text. The expository texts labeled as from ChatGPT were rated similarly to those labeled from Wikipedia on metrics of accuracy and well written-ness. However, the creative texts labeled as from ChatGPT were rated lower than texts labeled as from humans for buyability and creativeness. This may indicate that people comparatively believe ChatGPT is better at generating expository texts than creative texts. While this may initially seem to be surprising, there is still much that remains a mystery of how ChatGPT actually functions, and this can lead people to have an overestimation of its capabilities. If people do not have a general idea about how ChatGPT produces content, this could lead to difficulties in assessing the veracity of information from ChatGPT. In this new era of text generation, we need more transparency of text that is AI-generated as well as on how these systems work so people can make reasonable judgements on text they are told is AI-generated text. This is of utmost importance to clarify-as the line distinguishing what is written by humans and what is generated by AI continues to blur.

English

ISBN: 9798382305486Subjects--Topical Terms:

557829
Linguistics.
Subjects--Index Terms:

ChatGPT

From Wiki or ChatGPT? Perceptions of AI-Generated Text /
LDR:03561nam a22004333i 4500 001 1157773
005 20250603111413.5
006 m o d
007 cr|nu||||||||
008 250804s2024 miu||||||m |||||||eng d
020 $a 9798382305486
035 $a (MiAaPQD)AAI31238467
035 $a AAI31238467
040 $a MiAaPQD $b eng $c MiAaPQD $e rda
100 1 $a O'Boyle, Brianna Elizabeth, $e author. $3 1484042
245 1 0 $a From Wiki or ChatGPT? Perceptions of AI-Generated Text / $c Brianna Elizabeth O'Boyle.
264 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2024
300 $a 1 electronic resource (50 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Masters Abstracts International, Volume: 85-10.
500 $a Advisors: Dinkin, Aaron Committee members: Doyle, Gabriel; Reinholtz, Daniel.
502 $b M.A. $c San Diego State University $d 2024.
520 $a Arguably, one of the most revolutionary technological advancements of the 2020s has been the creation of ChatGPT, an AI-powered chatbot aided by innovations in the field of Natural Language Processing and the development of Large Language Models. There have been many people who use ChatGPT for its ability to generate text within seconds, but there remain questions of how do humans perceive AI-generated text, and are there certain genres of text people believe ChatGPT is better at generating. This study investigates human perception of AI-generated text using methodological frameworks from Sociolinguistics, specifically with a matched guise task. Participants were shown ten written texts of two genres, five expository texts and five creative texts, and asked to rate the texts based on three statements using a Likert scale. Participants were told the texts came from a variety of sources, but all of them were generated by ChatGPT. It was found in the majority of models participants were sensitive to the label of the text. The expository texts labeled as from ChatGPT were rated similarly to those labeled from Wikipedia on metrics of accuracy and well written-ness. However, the creative texts labeled as from ChatGPT were rated lower than texts labeled as from humans for buyability and creativeness. This may indicate that people comparatively believe ChatGPT is better at generating expository texts than creative texts. While this may initially seem to be surprising, there is still much that remains a mystery of how ChatGPT actually functions, and this can lead people to have an overestimation of its capabilities. If people do not have a general idea about how ChatGPT produces content, this could lead to difficulties in assessing the veracity of information from ChatGPT. In this new era of text generation, we need more transparency of text that is AI-generated as well as on how these systems work so people can make reasonable judgements on text they are told is AI-generated text. This is of utmost importance to clarify-as the line distinguishing what is written by humans and what is generated by AI continues to blur.
546 $a English
590 $a School code: 0220
650 4 $a Linguistics. $3 557829
650 4 $a Literature. $3 557269
650 4 $a Sociolinguistics. $3 555401
653 $a ChatGPT
653 $a AI-generated text
653 $a Natural Language Processing
653 $a Creative texts
690 $a 0290
690 $a 0401
690 $a 0636
690 $a 0800
710 2 $a San Diego State University. $b Linguistics. $3 1184077
720 1 $a Dinkin, Aaron $e degree supervisor.
773 0 $t Masters Abstracts International $g 85-10.
790 $a 0220
791 $a M.A.
792 $a 2024
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31238467