Effect of Cross-Model Stylistic and Behavioral Differences on User Perceptions

Investigate how stylistic and behavioral differences across large language models (e.g., tone and expression) affect users’ perceptions—such as self-efficacy and trust—during LLM-assisted writing.

Background

The study examined self-efficacy and trust but used a single LLM, limiting the ability to assess how different LLM styles or behaviors might shape user perceptions. Prior work suggests system tone and expression can influence these constructs, motivating a cross-model investigation.

The authors explicitly note they could not examine the effect of stylistic and behavioral variation across models, leaving open how such differences might change users’ perceived capabilities and reliance.

References

Our analysis also used a single LLM, so we could not examine how such stylistic and behavioral differences across models might alter user perceptions.

— Authorship Drift: How Self-Efficacy and Trust Evolve During LLM-Assisted Writing  (2602.05819 - Park et al., 5 Feb 2026) in Section 6 (Limitations and Future Work)

It remains unclear whether this reflects cognitive ease (editing a draft is more demanding than accepting it), trust in AI writing quality, the efficiency afforded by having a ready-made text, or simply the experimental setup.

— AI-Generated Email Drafts Shift Culturally Distinctive Communication Styles in Professional Email  (2609.26403 - Sakai et al., 22 Sep 2026) in Section 5, Limitations and Future Work