Get started
Work

An evaluation of AI-generated children's stories: Do word characteristics correlate with word learning?

Year: 2024
Type: dissertation
Abstract: Early childhood vocabulary acquisition has been extensively studied for its associations with future reading comprehension skills and academic performance. Shared storybook readings have been identified as a technique parents frequently use to expose children to new words. However, the word-related characteristics, word frequency, number of illustrations depicting each target word, and context variability have been recognized as variables that impact word learning from stories. Additionally, not all families have access to a repertoire of storybooks to read with their children, especially as they age and require stories with more challenging vocabulary. The current study investigated the use of artificial intelligence (AI) models for generating children’s stories to teach children new vocabulary. Word-related characteristics were also evaluated to determine if these factors correlated with learning outcomes. Thirty-two stories were generated using GPT-3, each containing five target words (OpenAI, 2020). Children between the ages of four and six read four stories with their parents before taking part in an elicited production task to test their knowledge of target words. Results indicated that children can learn new vocabulary words from AI-generated stories and that word frequency, number of illustrations, and context variability all positively correlated with word learning outcomes. Additionally, results highlighted a possible collinear relationship between the three word-related variables. These results inform our knowledge of early childhood word learning through AI-generated storybooks while also highlighting these AI models as a potential tool to aid vocabulary acquisition.
Author: McVay, Haley
Language: English
Field: Psychology
Open Access status: green
OpenAlex ID: W7000428245
Cites: 23
Cited by:
Related to: 10