AI as a Threat to Education? Contrasting GPT-3 and Google in Answering Questions Along Bloom’s Taxonomy of Educational Objectives
Nina Li*
York House School, 4176 Alexandra Street, British Columbia, Vancouver, Canada
*Corresponding Author: Nina Li, York House School, 4176 Alexandra Street, British Columbia, Vancouver, Canada.
Published: March 12, 2024
DOI: 10.55162/MCET.06.198
Abstract
Artificial Intelligence products have seen an unprecedented rise in the last few years, with GPT-3 and ChatGPT considered the most powerful AI tools in history. Recently, people have been concerned that services such as ChatGPT may be a threat to education. Specifically, teachers are concerned about students using the free and accessible tool as a Wikipedia replacement to complete homework and to write assignments for them. It poses importance for both researchers and practitioners in AI, and community stakeholders including teachers and parents, to understand the capability of AI text generative models in answering questions that may appear in an educational assessment or the school’s curriculum. In this research, we follow a widely adopted framework in defining educational question types, namely Bloom's taxonomy. We perform an experiment comparing the quality of answers retrieved from Google and GPT-3 on a series of questions identified along the hierarchy of Bloom’s taxonomy.
We summarize the capabilities and limitations of using GPT-3 to answer educational questions along Bloom's taxonomy. We discuss the implications on designing educational assessments to combat the threats introduced by such AI services.
Keywords: Generative language models; Question answering; AI in education; Bloom’s taxonomy
.