Document Type

Article

Department

Institute for Educational Development, East Africa

Abstract

The use of Generative Artificial Intelligence (AI) tools, such as Gemini and ChatGPT,  is  on  the  rise,  and  it  has  revolutionised  content  generation  in professional and academic domains. However, their increasing sophistication presents challenges in distinguishing AI-generated texts from human-written ones,  raising  concerns  about  integrity,  especially  in  academic  writing.  This study   evaluated   the   effectiveness   of   three   AI   text   humanising   tools: Writesonic, QuillBot paraphraser, and WriteHuman in refining texts generated by ChatGPT and  Gemini.  The study employed  a  comparative  experimental design.  By  employing  quantitative  analysis,  the  study  compares  baseline  AI detection    rates    with    those    after    successive    humanisation    iterations. Comparison  was  also  made  across  different  humanising  tools.  The  AI detecting  tools  that  were  used  in  the  study  are:  QuillBot,  ZeroGPT,  and Scribbr.  Findings  show  variations  in  tool  effectiveness.  While  WriteHuman achieved  consistent  success  in  masking  AI  text  origins  with  an  Average Detection   Rate   (ADR)   of   1.98%,   Writesonic   and   QuillBot   exhibited inconsistent   performances   in   their   humanising   processes   with   Average Detection Rates of 64.39% and 93.56%, respectively.  Across all iterations and AI detectors, Writesonic and QuillBot detection rates fluctuated significantly, sometimes  improving  and  then  regressing  unpredictably,  highlighting  that these tools are not reliable. Further, QuillBot performed poorly with its 15 out of 18 outputs from 18 iterations remaining flagged as AI-generatedin 15 out of  18  iterations  across  all  AI  text  detectors.  This  study  highlights  ethical implications,  with  emphasis  on  the  need  for  users  to  use  these  tools  with caution, since their effectiveness is not guaranteed. Further, policies governing the usage ofthese tools in academic writing should be put in place by academic institutions. This study makes a significant contribution to understanding the interplay between Generative AI tools, AI content detection technologies, and humanising strategies, fostering informed discourse on academic integrity in the era of AI

AKU Student

no

Publication (Name of Journal)

International Journal of Advanced Research

DOI

https://doi.org/10.37284/ijar.9.1.4683

Creative Commons License

Creative Commons Attribution-Share Alike 4.0 International License
This work is licensed under a Creative Commons Attribution-Share Alike 4.0 International License.

Download

Share

COinS