Authorship Drift: How Self-Efficacy and Trust Evolve During LLM-Assisted Writing

TL;DR

This paper investigates how self-efficacy and trust evolve during LLM-assisted writing, treating them as turn-level, task-specific states rather than static traits. Using an online study of 302 participants writing argumentative essays with an embedded LLM, the authors identify five self-efficacy trajectories and four trust trajectories, and analyze how prompting strategies relate to these dynamics and to actual versus perceived authorship. They find that self-efficacy generally declines while trust increases, with higher trust buffering declines; prompting patterns such as editing-heavy strategies correlate with reduced authorship, whereas review-oriented interactions support authorship recovery. The work contributes to understanding authorship in human–AI collaboration and offers design implications to support user agency, calibration of trust, and visibility of collaboration history in AI-assisted writing.”,

Abstract

Large language models (LLMs) are increasingly used as collaborative partners in writing. However, this raises a critical challenge of authorship, as users and models jointly shape text across interaction turns. Understanding authorship in this context requires examining users' evolving internal states during collaboration, particularly self-efficacy and trust. Yet, the dynamics of these states and their associations with users' prompting strategies and authorship outcomes remain underexplored. We examined these dynamics through a study of 302 participants in LLM-assisted writing, capturing interaction logs and turn-by-turn self-efficacy and trust ratings. Our analysis showed that collaboration generally decreased users' self-efficacy while increasing trust. Participants who lost self-efficacy were more likely to ask the LLM to edit their work directly, whereas those who recovered self-efficacy requested more review and feedback. Furthermore, participants with stable self-efficacy showed higher actual and perceived authorship of the final text. Based on these findings, we propose design implications for understanding and supporting authorship in human-LLM collaboration.

Figures (8)

• Figure 1: Overview of the study interface: (a) LLM-assisted writing workspace and (b) real-time self-efficacy and trust rating
• Figure 2: Normalized overlays of self-efficacy and trust trajectories for the five major trajectory patterns. The horizontal axis shows the time-normalized turn-by-turn progression from initial pre-survey rating (left) to final rating (right). The vertical axis shows the relative change in Likert-scale points after zero-centering all trajectories at each participant's pre-survey rating. Thin translucent lines represent individual participants, the solid line shows the group median, and the shaded band indicates the interquartile range.
• Figure 3: Proportions of editing and reviewing prompts across different self-efficacy trajectory patterns. (*$p<0.05$, ***$p<0.001$)
• Figure 4: Proportions of information searching prompts across different trust trajectory patterns. (*$p<0.05$)
• Figure 5: Proportion of drafting---editing and editing---editing transition across different self-efficacy trajectory patterns. (**$p<0.01$, ***$p<0.001$)