Systematic prompting of an LLM can generate cognitively activating formative feedback aligned with research-based scaffold types (demonstrated, not yet empirically validated)¶

Claim · All claims
Evidence · 1 study · design r1 · q2
Currency · newest study 2024 · checked against its sources 2026-10-19 (automated, 1 of 1 entry) · not reviewed by a person · recheck by 2027: tool and model accuracy changes quickly

Subclaims¶

q2 i? In a demonstration with a Newton's cradle task, strategic prompting of GPT-4 turbo produced feedback formatted as cognitively activating instructions covering metacognitive, self-explanatory, and motivational scaffolds. → Steinert 2024

Evidence¶

Steinert 2024¶

Steinert, S., Avila, K. E., Ruzika, S., Kuhn, J., & Küchemann, S. (2024). Harnessing large language models to develop research-based learning assistants for formative feedback. Smart Learning Environments. https://doi.org/10.1186/s40561-024-00354-1

q2 · i? · design · r1

The authors' demonstration of LEAP on a Newton's cradle video task, with prompts and GPT-4 turbo responses shown in Table 1. The article notes the scaffolds could "potentially assist students in various self-regulating processes"; no learning outcomes were measured.

"The feedback provided by the LLM, as shown in Table 1, demonstrate how strategic prompting can be formatted as cognitively activating instructions to potentially assist students in various self-regulating processes."

Discussion¶

Related Claims¶

• Scaffolding improves learning — a broader claim this one bears on
• Phase VI AI-enabled systems face six documented challenges: explainability, hallucination risk, prompt sensitivity, computational cost, validation complexity, and limited large-scale evidence — related
• Reviewed literature reports risks of overreliance, AI errors in complex tasks, and students' uncritical acceptance of AI-generated outputs — related
• Despite generally accurate feedback, the LLM's scoring explanations could be unpredictable and prone to logical inconsistency, failing to award points even when citing the correct rubric directive — related
• More complex prompts tend to elicit superior results with advanced LLM versions (authors' observation) — related
• Human-LLM agreement remains low across constructs (κ = 0.000 to 0.419), with self-efficacy showing the best alignment and Prior KSAs none — reports the opposite
• Self-regulated learning improves achievement — related
• Instructor-scripted self-explanation prompts have limitations, motivating computer-generated prompts — related