Investigating Middle School Students Question-Asking and Answer-Evaluation Skills When Using ChatGPT for Science Investigation
Project Overview
The document explores the integration of generative AI tools, such as ChatGPT, in middle school science education, particularly examining their impact on students' abilities to ask questions and evaluate answers. The study indicates that while there is a growing utilization of these AI resources by students, many face challenges in crafting effective questions and critically analyzing AI-generated responses. This over-reliance on AI answers has resulted in only moderate learning outcomes, suggesting that access to advanced AI tools alone is insufficient for enhancing educational achievement. Consequently, the findings underscore the necessity for targeted educational interventions to bolster students' AI literacy and improve their question-asking skills, ultimately aiming to maximize the benefits of generative AI in the learning environment.
Key Applications
ChatGPT
Context: Middle school science investigation tasks
Implementation: Students used ChatGPT to solve science problems by generating prompts and evaluating responses.
Outcomes: Students demonstrated limited ability to ask effective questions and evaluate AI responses; learning outcomes were moderate.
Challenges: Students often over-relied on AI, struggled with formulating clear questions, and failed to critically evaluate the answers.
Implementation Barriers
Skill-related / Over-reliance
Students lack effective question-asking and answer-evaluation skills when using AI and tend to accept AI responses without critical evaluation, leading to misconceptions.
Proposed Solutions: Educational interventions promoting AI literacy, scaffolding metacognitive processes, and training to develop critical thinking and awareness of AI limitations.
Project Team
Rania Abdelghani
Researcher
Kou Murayama
Researcher
Celeste Kidd
Researcher
Hélène Sauzéon
Researcher
Pierre-Yves Oudeyer
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Rania Abdelghani, Kou Murayama, Celeste Kidd, Hélène Sauzéon, Pierre-Yves Oudeyer
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