Generative AI and AI-generated Contents on Social Media

Permanent URI for this collectionhttps://hdl.handle.net/10125/107458

Browse

Recent Submissions

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item type: Item ,
    Incorporating Artificial Intelligence into Student Academic Writing in Higher Education: The Use of Wordtune by Chinese international students
    (2024-01-03) Zhao, Xin; Xu , Jiahong; Cox, Andrew
    Academic writing can be challenging for international students, especially if English is not their first language. Artificial intelligence (AI) writing assistants have received considerable attention in recent years as a new means to enhance students’ academic writing. However, limited research has been conducted on how they are actually used in practice. This paper examines the use of Wordtune, an AI-powered writing assistant, by Chinese international students in higher education through interviews (n=30). The study explored the challenges these students faced in academic writing and how they already used a variety of digital tools during the writing process to address these issues. Specifically in relation to Wordtune students found the rewriting options useful, especially the function to rewrite in formal language. Students self-identifying as beginners in English used all the functions, but rather indiscriminately. Students with higher-level skills used it more selectively and learned to improve their writing through examining alternative rewrites. All users wanted the function to rewrite sentences more formally to suit an academic writing style. The paper contributes to our understanding of how international students use digital tools in the writing process.
  • Thumbnail Image
    Item type: Item ,
    How good can machine generated texts be identified and can language models be trained to avoid identification?
    (2024-01-03) Schneider, Sinclair; Steuber, Florian; Schneider, João A. G.; Dreo Rodosek, Gabi
    With the rise of generative pre-trained transformer models such as GPT-3, GPT-NeoX, or OPT, distinguishing human-generated texts from machine-generated ones has become important. We refined five separate language models to generate synthetic tweets, uncovering that shallow learning classification algorithms, like Naive Bayes, achieve detection accuracy between 0.6 and 0.8. Shallow learning classifiers differ from human-based detection, especially when using higher temperature values during text generation, resulting in a lower detection rate. Humans prioritize linguistic acceptability, which tends to be higher at lower temperature values. In contrast, transformer-based classifiers have an accuracy of 0.9 and above. We found that using a reinforcement learning approach to refine our generative models can successfully evade BERT-based classifiers with a detection accuracy of 0.15 or less.
  • Thumbnail Image
    Item type: Item ,