Framing, Judging, Steering: An Assessable Competency Model for Teach-ing Students to Reason With Generative AI

Researchers propose CoRe-3, a three-part competency model for teaching students to reason effectively with generative AI by separating task framing, output evaluation, and iterative steering into distinct, measurable skills. The framework addresses a critical gap in AI education: current assessments collapse productive AI use into a single 'prompting' score, obscuring where students succeed or fail in working with AI systems.

Educational institutions face a fundamental mismatch between how they assess student work and how students actually use AI in practice. Traditional metrics measure unaided performance, yet the real-world skill involves collaborating with AI to produce quality output through deliberate problem-solving stages. The CoRe-3 framework (Framing, Judging, Steering) tackles this by decomposing AI collaboration into three distinct competencies, each with independent assessment pathways.

The research builds on growing recognition that AI literacy requires more than prompt-writing tutorials. Framing—articulating ill-defined problems before querying AI—demands deep subject knowledge and metacognitive clarity. Judging requires critical evaluation of AI output against unstated assumptions and domain-specific criteria. Steering involves iterative refinement through targeted feedback loops. By treating these as separate skills, educators gain diagnostic precision impossible with aggregate 'prompting' scores.

The instantiation in CoReasoningLab demonstrates technical rigor: using simulated learners across different AI backends proves the skills dissociate meaningfully (discriminant validity) while remaining interdependent (convergent validity). This cross-model validation suggests the framework transcends vendor-specific quirks. The approach directly addresses AI's cognitive offloading risk—students can appear productive while developing shallow reasoning habits.

The framework's release as open-source infrastructure positions it as foundational for AI competency assessment in schools. As institutions race to integrate AI into curricula, validated measurement tools become critical gatekeeping mechanisms. Success here could shift pedagogical focus from tool mastery to reasoning architecture, with downstream implications for workforce readiness and AI governance literacy.

• →CoRe-3 separates AI collaboration into three assessable skills—framing problems, judging outputs, and steering iterations—replacing vague 'prompting' evaluations with diagnostic precision.
• →The model addresses cognitive offloading risks by forcing explicit reasoning stages before and after AI generation, preventing students from outsourcing thinking.
• →Cross-model validation across different AI backends suggests the framework transcends vendor specificity and could become a standard assessment tool.
• →Educational institutions now have an open-source platform to measure AI reasoning competency, enabling data-driven curriculum design rather than anecdotal integration.
• →The research implies future AI literacy benchmarks will prioritize metacognitive frameworks over tool familiarity, shifting workforce preparation priorities.