Generative AI and technological mediation: writing micro-processes, products and perceptions

Abstract

The public release of ChatGPT presents language education with both advantages and challenges. Being widely available and readily producing texts at user's command, it is an especially useful resource in ESL writing. However, to be able to leverage it effectively for that purpose, we need to understand the writing micro-processes surrounding its use. To this end, this study investigates, through an experimental design, writing micro-processes along with writing outcomes, and perceptions of 12 ESL writers across three writing conditions. The data included interviews, questionnaires as well as essay grades and screen recordings. Two of the three conditions involved the use of ChatGPT. The results indicate that the majority of participants favored the ChatGPT- assisted conditions and noted its role in aiding idea generation. Despite this preference, some participants noted that ChatGPT-produced texts felt robotic and lacked a human touch. We found no statistically significant differences across the three conditions in the participants' essays. Most notably, our analysis identified distinct clusters of writing micro-processes associated with the ChatGPT conditions, suggesting that generative AI tools can influence writing behaviors although the link to outcomes is not straightforward. These findings provide insights into the potential benefits and limitations of using AI tools in writing tasks. The article discusses areas for further research to enhance the integration of AI into ESL writing education.