The Value, Benefits, and Concerns of Generative AI-Powered Assistance in Writing

TL;DR

The results suggest that people are willing to forgo financial payments to receive writing assistance from AI, especially if AI can provide direct content generation assistance and the writing task is highly creative.

Abstract

Recent advances in generative AI technologies like large language models raise both excitement and concerns about the future of human-AI co-creation in writing. To unpack people's attitude towards and experience with generative AI-powered writing assistants, in this paper, we conduct an experiment to understand whether and how much value people attach to AI assistance, and how the incorporation of AI assistance in writing workflows changes people's writing perceptions and performance. Our results suggest that people are willing to forgo financial payments to receive writing assistance from AI, especially if AI can provide direct content generation assistance and the writing task is highly creative. Generative AI-powered assistance is found to offer benefits in increasing people's productivity and confidence in writing. However, direct content generation assistance offered by AI also comes with risks, including decreasing people's sense of accountability and diversity in writing. We conclude by discussing the implications of our findings.

Figures (13)

• Figure 1: Comparing participants' probability of preferring an AI-assisted writing mode over the independent writing mode, (a) when writing either argumentative essays or creative stories, (b) when writing argumentative essays (arguments) only, or (c) when writing creative stories only. The wage premium is defined as the difference in the writing job payment participants would receive from selecting the AI-assisted writing mode, relative to the fixed payment of $3 from the independent writing mode. Triangles (or squares) represent the actual fraction of participants who chose the AI-primary (or human-primary) writing mode over the independent writing mode in our dataset. The blue solid line (or the orange dotted line) depicts the predicted probability, derived from the fitted probit model, of participants opting for the AI-primary (or human-primary) writing mode over the independent writing mode. The AI assistance's value is the negative of the wage premium for the AI-assisted job that makes participants indifferent between the AI-assisted mode and the independent writing mode (i.e., choosing the AI-assisted writing mode with a probability of 0.5), which is marked by pentagrams.
• Figure 2: Comparing the estimated value of AI assistance between participants with varying levels of writing confidence for different types of writing tasks.
• Figure 3: Comparing the estimated value of AI assistance between participants with varying levels of familiarity with ChatGPT for different types of writing tasks.
• Figure 4: Comparing the predicted cognitive load that the average participant experienced for completing the two types of writing tasks (argumentative essay writing and creative story writing), across the independent, human-primary, and AI-primary writing modes. Error bars represent the 95% confidence intervals of the predicted cognitive load level. $\textsuperscript{*}$ denotes statistical significance levels of $0.05$.
• Figure 5: Comparing the predicted perceptions of the average participant regarding the overall writing process for completing the two types of writing tasks (argumentative essay writing and creative story writing), across the independent, human-primary, and AI-primary writing modes. Error bars represent the 95% confidence intervals. $\textsuperscript{*}$ and $\textsuperscript{**}$ denote statistical significance levels of $0.05$ and $0.01$, respectively.