Our Coding Adventure: Using LLMs to Personalise the Narrative of a Tangible Programming Robot for Preschoolers
Project Overview
The document explores the application of Large Language Models (LLMs) in preschool education, particularly in crafting personalized storytelling for programming robots like Cubetto. This innovative approach enables preschool teachers to create engaging narratives that aid young learners in grasping programming concepts indirectly. By leveraging LLMs, educators can enhance children's computational thinking skills through storytelling, making complex ideas accessible and enjoyable. However, the implementation of LLMs is not without challenges, as issues related to output consistency and hallucinations can arise. Despite these obstacles, the potential benefits of using generative AI in educational settings, particularly in fostering creativity and understanding in early learners, underscore its promise as a tool for enhancing instructional practices in preschool environments.
Key Applications
Using LLMs to generate personalized storytelling for Cubetto, a tangible programming robot.
Context: Preschool education, targeting preschool children and their teachers.
Implementation: Teachers use LLMs to generate narratives based on children's interests, which are then utilized to program the Cubetto robot.
Outcomes: Enhanced engagement and understanding of programming concepts for preschoolers, with narratives tailored to their preferences.
Challenges: Inconsistencies in LLM outputs, including hallucinations and incomplete instructions.
Implementation Barriers
Technical Barrier
Issues with LLM outputs, such as hallucinations and inconsistencies in generated narratives.
Proposed Solutions: Teachers can provide explicit prompts and guidelines to improve LLM output quality and ensure it fits the educational context.
Implementation Barrier
The need for teacher training to effectively utilize LLMs without exposing children to them directly.
Proposed Solutions: Provide professional development and resources for teachers to enhance their understanding of LLMs and proper implementation in classrooms.
Project Team
Martin Ruskov
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Martin Ruskov
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai