Objective: This study aims to investigate how using generative AI such as ChatGPT influences users' cognitive dependency that results in cognitive offloading and deskilling.

Background: Generative AI, specifically LLM (Large Language Model)'s use had been rapidly growing in the creative domains such as writing. Some studies suggest that using LLM, or mainly, ChatGPT can support idea generation and expedite learnings, while others warn that users may experience deskilling and weakening on critical thinking. Most of the discussions on this topic have approached from the educational perspective focusing primarily on the ethical implications or the quality of outcome. On the other hand, relatively little attention has been paid to the process of using LLM and how human cognitions are affected.

Method: A total of 64 creative essays were collected from 32 University students. These students either used or not used ChatGPT to write the essays. Three writing experts (who identifies their job as a "writer" and has more than 5 years of writing experience) rated the essays on the number of assertions, the number of opposing views, the number of detailed cases introduced, concept quality, sense of agency, and structure diversity. For analysis, the study saw the difference between ChatGPT using group and no support group using paired-sample t-test.

Results: The results showed statistically significant differences between AI vs non- AI assisted groups. Participants who used ChatGPT during writing had used more counterclaims (p<0.001), while participants who wrote without ChatGPT demonstrated significant higher levels of authorial agency (p<0.001). Some marginal trends were found in the logical complexity, showing higher number of case use and complexity for the non-AI assisted group.

Conclusion: This study suggests that while generative AI can support humans, it may also potentially reduce and deskill individual cognitive abilities when articulating their arguments. In conclusion, this study suggests the need for generative AI systems from human factors perspective, so that the human-AI interaction helps support productivity while maintaining users' cognitive engagement and human agency.

Application: The results of this study may contribute to the development of AI interfaces and human-AI interaction design (AX), that support productivity while reducing deskilling and cognitive offloading.