Abstract:

In the past year alone, Large Language Models (LLMs) like ChatGPT, Bard, and GPT-4 have transformed from theoretical conceptions about artificial intelligence (AI) to accessible, free, and widely available tools. With generative AI technologies dominating international discourse and maintaining upward trajectories in both prevalence and capability, industries across the board are working to understand the potential uses of LLMs to transform the way things get done.

The emergence of these technologies occurred as educational systems are still grappling with lasting effects of the global pandemic and the extent of covid-related learning loss is on the forefront of educators' priorities. Meanwhile, emerging LLMs offer promising opportunities to students, teachers, and administrators: conversational AI tools can talk students through concepts, offer constructive feedback on writing, generate lesson plans and instructional materials, and more. Yet, the risks are conspicuous: LLMs enable students to complete assignments without fully understanding content material, and given the mechanics of generative AI, detecting LLM use is complicated. Given the novelty of these technologies, there is not yet a common understanding of what is considered plagiarism, how students should be directed (or even encouraged) to use AI-based tools, and what the long-term outlooks are for education and the workforce.

This paper is the first in a series of research to synthesize a unifying framework for generative AI in K-12 education. Herein, we explore these ideas, among others, in more detail, offering perspectives from early literature in this space and insights from a series of interviews conducted with teachers and administrators across the state of New Jersey. Implications for research and practice are discussed.

Keywords:

Large language models, artificial intelligence, education.