Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

How Does the Disclosure of AI Assistance Affect the Perceptions of Writing?

Zhuoyan Li, Chen Liang, Jing Peng, Ming Yin

TL;DR

The paper investigates how disclosing AI involvement during writing affects reader perceptions of quality. It uses a two-phase experimental design with argumentative essays and creative stories, manipulating AI involvement across Independent, AI editing, and AI content-generation modes. Key findings show that disclosure—especially when AI contributes content—reduces perceived quality and increases cross-rater variability, with effects moderated by writers’ confidence and familiarity with AI tools. The work has practical implications for labeling AI-assisted content on platforms and emphasizes the need for transparent, ethically informed disclosure strategies in human-AI co-authored writing.

Abstract

Recent advances in generative AI technologies like large language models have boosted the incorporation of AI assistance in writing workflows, leading to the rise of a new paradigm of human-AI co-creation in writing. To understand how people perceive writings that are produced under this paradigm, in this paper, we conduct an experimental study to understand whether and how the disclosure of the level and type of AI assistance in the writing process would affect people's perceptions of the writing on various aspects, including their evaluation on the quality of the writing and their ranking of different writings. Our results suggest that disclosing the AI assistance in the writing process, especially if AI has provided assistance in generating new content, decreases the average quality ratings for both argumentative essays and creative stories. This decrease in the average quality ratings often comes with an increased level of variations in different individuals' quality evaluations of the same writing. Indeed, factors such as an individual's writing confidence and familiarity with AI writing assistants are shown to moderate the impact of AI assistance disclosure on their writing quality evaluations. We also find that disclosing the use of AI assistance may significantly reduce the proportion of writings produced with AI's content generation assistance among the top-ranked writings.

How Does the Disclosure of AI Assistance Affect the Perceptions of Writing?

TL;DR

The paper investigates how disclosing AI involvement during writing affects reader perceptions of quality. It uses a two-phase experimental design with argumentative essays and creative stories, manipulating AI involvement across Independent, AI editing, and AI content-generation modes. Key findings show that disclosure—especially when AI contributes content—reduces perceived quality and increases cross-rater variability, with effects moderated by writers’ confidence and familiarity with AI tools. The work has practical implications for labeling AI-assisted content on platforms and emphasizes the need for transparent, ethically informed disclosure strategies in human-AI co-authored writing.

Abstract

Recent advances in generative AI technologies like large language models have boosted the incorporation of AI assistance in writing workflows, leading to the rise of a new paradigm of human-AI co-creation in writing. To understand how people perceive writings that are produced under this paradigm, in this paper, we conduct an experimental study to understand whether and how the disclosure of the level and type of AI assistance in the writing process would affect people's perceptions of the writing on various aspects, including their evaluation on the quality of the writing and their ranking of different writings. Our results suggest that disclosing the AI assistance in the writing process, especially if AI has provided assistance in generating new content, decreases the average quality ratings for both argumentative essays and creative stories. This decrease in the average quality ratings often comes with an increased level of variations in different individuals' quality evaluations of the same writing. Indeed, factors such as an individual's writing confidence and familiarity with AI writing assistants are shown to moderate the impact of AI assistance disclosure on their writing quality evaluations. We also find that disclosing the use of AI assistance may significantly reduce the proportion of writings produced with AI's content generation assistance among the top-ranked writings.
Paper Structure (38 sections, 30 figures, 3 tables)

This paper contains 38 sections, 30 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Assisting human writing using large language models.
  4. Human-centered evaluations of texts generated or edited by large language models.
  5. Study Design
  6. Phase 1: Collection of Writing Samples
  7. Writing tasks.
  8. Writing modes.
  9. Procedure.
  10. Data Collection Results.
  11. Phase 2: Experimental Design
  12. Phase 2: Experimental Procedure
  13. Background assessment.
  14. Main rating tasks.
  15. Exit Survey.
  16. ...and 23 more sections

Figures (30)

  • Figure 1: Comparing average ratings of the overall quality of articles generated under the independent, AI editing, or AI generation writing modes, with and without disclosure of the use and type of AI assistance during the writing process. Error bars represent the 95% confidence intervals of the mean values. $\textsuperscript{*}$ and $\textsuperscript{***}$ denote significance levels of $0.05$ and $0.001$, respectively.
  • Figure 2: Comparing the variance in the overall quality ratings of articles generated under the independent, AI editing, or AI generation writing modes, with and without disclosure of the use and type of AI assistance during the writing process. Error bars represent the 95% confidence intervals of the variance. $\textsuperscript{**}$ denotes the significance level of $0.01$.
  • Figure 3: The average difference between an article's overall quality ratings in the " Disclose" and " Non-Disclose" treatments, among raters with high versus low confidence in their own writing skills. Error bars represent the 95% bootstrap confidence intervals of the rating difference. An interval below zero means the corresponding group of raters significantly decrease their ratings when the use and type of AI assistance in the writing process was revealed to them.
  • Figure 4: The average difference between an article's overall quality ratings in the " Disclose" and " Non-Disclose" treatments, among raters with high versus low familiarity with ChatGPT. Error bars represent the 95% bootstrap confidence intervals of the rating difference. An interval below zero means the corresponding group of raters significantly decrease their ratings when the use and type of AI assistance in the writing process was revealed to them.
  • Figure 5: Within the top $\gamma \%$ of articles for the same writing task (ranked by articles' average overall quality ratings), the percentages of articles that were written in each of the three writing modes, with and without disclosing the use and type of AI assistance. $\textsuperscript{**}$ and $\textsuperscript{***}$ denote the significance level of $0.01$ and $0.001$, respectively.
  • ...and 25 more figures