Speakers
Description
Written responses generated by Artificial Intelligence (AI) has been commonly used by language learners with a desire to better visualize and interpret band descriptors of examinations. However, the concerns raised against its ability to create sample responses that reflect the complexity and diversity of coherence and cohesion in human-written responses. By adopting the theoretical framework of Systemic Functional Grammar (SFG) to analyze the thematic choice, the research aimed to examine the differences in the organization of the clause and text, and compare thematic progression patterns on VSTEP essays written by human and Gen AI. The finding suggests that humans tend to employ more varied interpersonal themes and marked themes for making emphasis while Gen AI can overuse textual themes to list ideas. It is also more common for humans to use linear progression whereas Gen AI relies mostly on constant theme progression, making the text cohesive yet repetitive and monotonous.
Key words: VSTEP, Systemic Functional Grammar, thematic choice, thematic progression pattern
