We are facing a future in which artificial intelligence (AI) is capable of producing accurate and sophisticated responses to prompts by generating text, images, multimedia experiences, and software. Even today’s K–12 students, with existing Generative AI (GenAI) chatbots, will likely spend significantly less time creating these products by the time they enter the workforce. However, GenAI is still just a tool, albeit an increasingly powerful one. It relies on thoughtful human use. Humans will still need to define what AI should produce and how, and what it should not and why. They will also need to evaluate the product—not just for technical quality, but also for how well the product represents them and meets human needs. These are questions that people have already been wrestling with, but they are especially important for students to understand and incorporate into practice. Teaching and assessing students in this future where craft and creation are redefined has already proven challenging. We need a fundamental shift in what knowledge and skills are prioritized in the curriculum and how they are sequenced and scaffolded. Critical, creative, and metacognitive skills, which are often sidelined in curricula, will need to be centered to drive ethical and productive use of AI. Many organizations worldwide have charted these substantial shifts and developed guidelines for teaching students what AI is and how to work with it responsibly, critically, and effectively. One common skill set that these frameworks emphasize is problem solving. However, they do not adequately address one essential element of problem solving: how students will identify meaningful problems to solve.

We conducted a literature review with the goal of discovering teachable skills that would become increasingly relevant assuming that AI will become accurate and efficient at problem-solving. In particular, we wanted to identify psychological skills that were supported by a strong empirical research base, including research on developmental trajectories, pedagogical approaches, and methods of assessment. We conducted this review in two phases.

The first phase constitutes a scoping review to address the following research questions:
1) What psychological skill(s) will prepare students to direct GenAI tools to solve meaningful and important problems?
2) How have researchers already defined these skills?
3) What is the depth and breadth of research on these skills?

In the second phase, we conducted a systematic literature review on problem finding, which our scoping review found had a strong empirical research base but also lacked a recent literature review.

Our research questions for the second phase were:
1) How does problem finding skill develop?
2) How do we teach problem finding?
3) How do we measure problem finding skill?

We also reviewed K-12 academic standards and frameworks used in the United States to understand how problem-finding may already be implemented in curriculum. In this paper, we will summarize our process and findings from the literature review and make recommendations for research and curriculum.

Keywords: Curriculum, generative artificial intelligence, K-12 education, instruction, large language models, problem finding, teaching.