Investigating ChatGPT's Performance in Students' Academic Writing

• Main Authors: Esmas, Robian M., Florendo, Peter Ivan A., Somontina, Joellianne Grace B., Malantic, Diana Nicole F., Beredo, Jackylyn L.
• Format: text
• Published: Animo Repository 2024
• Subjects: Artificial Intelligence; Generative AI; ChatGPT; academic writing; students
• Online Access: https://animorepository.dlsu.edu.ph/conf_shsrescon/2024/paper_csr/6; https://animorepository.dlsu.edu.ph/context/conf_shsrescon/article/2413/viewcontent/PP_CSR_Esmas_Florendo_Somontina_Malantic_Beredo___Robian_Esmas.docx.pdf
• Institution: De La Salle University

As Artificial Intelligence (AI) continues to evolve, its integration into various domains has become increasingly notable. One prominent example is ChatGPT, a software tool that has gained popularity as an academic writing assistant. Despite its popularity, ChatGPT has sparked controversies due to concerns about potential misuse and exploitation. Studies have identified several writing challenges students face, especially those learning English as a foreign language (EFL), including grammar and spelling, vocabulary, paraphrasing, and referencing. This study evaluates ChatGPT 3.5's effectiveness in addressing these specific writing skills. Participants composed of Senior High School (SHS) students were assigned an academic writing task with the freedom to use ChatGPT. Each participant’s conversation logs were collected and analyzed using Grammarly, Text Inspector, Turnitin, and Scribbr. The results indicate that while ChatGPT is a useful writing assistant, it has notable limitations in areas including complex sentence construction, excessive formality, and the generation of non-existent links, which may vary depending on the prompt used and its length. By highlighting these assessments, our study contributes to the ongoing discussions about potential applications and limitations of ChatGPT in academic writing and education.