Idea Generation Strategies for Student Writing with AI-NLG Tools

Overview

Idea generation is fundamental to writing, but students often struggle with the idea generation in classroom writing contexts. Artificial intelligence (AI) natural language generation tools that produce coherent texts based on input text might help address this problem.

In this seminar, the speaker will explore the strategies by which students interact with these NLG tools to write short stories. The school of this speaker has grounded this study in data collected from the first Human-AI Creative Writing Contest in a mainstream Hong Kong secondary school. In analysing students' reflections, the speaker found different patterns by which students produce input text for NLG tools, select output text from NLG tools, and decide on particular NLG tools to use. In analysing the language features of students' short stories, we found different patterns by which students integrated words from NLG tools into the stories. Besides, human raters scored the stories, and some strategies for writing with AI-NLG tools appeared to be more effective for human ratings. As NLG tools are rarely deployed for writing activities in classroom contexts, our presentation can inform how to design NLG tools for classroom writing, and how teachers might guide students to effectively use NLG tools for idea generation.