Mutual critique is essential in university learning, as it helps students revise their ideas and proposals. However, implementing peer critique among students is challenging. Students often perceive criticism as a personal attack, which leads to fear or aversion. This fear mixed with empathy also discourages them from criticizing others. Thus, ensuring psychological safety is crucial for promoting effective peer critique. To this end, we propose a “Sense of Ownership Reduction Technique” using generative AI for style transformation. In this method, students’ texts are rewritten by AI in a different style. Peer critiques are conducted on the AI-transformed versions, and the students revise these based on the feedback. Finally, the revised texts are retransformed into the original style of each student. Through separation of the personal style from the content, students feel less ownership over the text, which may reduce the discomfort associated with criticism.

This study reports a foundational investigation aimed at establishing an effective implementation method for the proposed technique. It focuses on two key questions:
(1) How does style transformation affect the sense of ownership over the text?
(2) To what extent does it reduce discomfort toward different types of criticism?

In the study, 191 university students first wrote an opinion essay about an Australian ban on social media use by children under 16 years of age. Each text was then transformed in four ways using ChatGPT:
a) proofreading (minor corrections),
b) academic style,
c) casual style, and
d) evidence-enhanced version (AI adds justifications). Participants rated each version on:
(1) sense of ownership and
(2) discomfort toward different types of critique.

They ranked ownership highest for the original and proofread versions, moderate for the academic and casual styles, and lowest for the evidence-enhanced version. Discomfort toward criticism of logic and persuasiveness was also lower for the academic and casual styles than for the original style. The evidence-enhanced version was associated with the least discomfort.

The evidence-enhanced version elicited both a low sense of ownership and low discomfort in response to criticism. However, because it diverged significantly from the original text, students were less likely to perceive the revision process as personally meaningful. Therefore, transforming the text into either an academic or casual style appears to be a more appropriate method to reduce resistance to criticism while encouraging active engagement in revision. Overall, academic and casual style transformations can be effective in lowering emotional resistance to critique while still allowing for meaningful revisions. Future work will refine this method through further analyses and additional experiments.

Keywords: Mutual criticism, Psychological safety, AI- based style transformation, sense of ownership reduction technique.