Design GenAI learning support to withhold direct answers and prompt students to articulate reasoning and generate solution ideas¶

Strategy · All strategies
Evidence · 5 claims (5 for) · 6 studies (2 causal, 2 theoretical, 1 review, 1 qualitative), q1–q3 · 0 of 6 report an effect size · 4 claims rest on one study

Description¶

In this RCT, all GPTs across conditions were given a common design feature: "All GPTs were explicitly instructed not to provide correct answers, even if requested". The authors justify this by evidence that students often request complete solutions prematurely and that direct answer provision can reduce generative cognitive engagement, sometimes described as "metacognitive laziness". The design deliberately prioritized cognitive activation by prompting students to articulate reasoning and generate solution ideas, with correct solutions provided on a subsequent page for corrective feedback and self-correction.

Design Implications¶

Context¶

Requirements¶

• Prompts must be developed through iterative prompt-engineering with a structured knowledge base covering tasks, solutions, and intervention principles; GPTs need access to task materials and solutions to reduce hallucinations

Constraints¶

• Withholding solutions may appear pedagogically inauthentic, as the authors acknowledge; the approach targets SRL processes rather than optimizing performance or providing direct tutoring through answer provision

Target Learners¶

• secondary school students (Grades 7–9) in regular physics and English lessons

Target Learning Goals¶

• cognitive activation, generative engagement with learning tasks, and self-regulated learning processes

Claims¶

• Neither GenAI intervention outperformed standard ChatGPT control on effort, domain-specific knowledge, or elaboration-based strategy use [+W]
• GenAI-supported utility value reflection produced more favorable development of perceived utility value than GenAI-supported cognitive learning strategy prompting in secondary students [+W]
• Students who engaged more meaningfully with the GenAI tended to have more sustained interest than control students (exploratory) [+W]
• Reviews citing one randomised trial (Bastani et al. 2025) report that unrestricted GPT-4 access in high-school mathematics raised assisted practice performance but left students worse off after withdrawal [+W]
• Greater answer offloading during learning is associated with lower subsequent unaided performance (exploratory) [+W]

Related Strategies¶

• Reasoning-Focused Instruction

Examples¶

-

Key Sources¶

• Tim Fütterer, Lisa Bardach, Jochen Kuhn, Stefan Daniel Keller, Peter Gerjets. (2026). Enhancing School Students’ Self-Regulated Learning through Generative AI Support: A Randomized Controlled Trial. Educational Psychology Review. https://doi.org/10.1007/s10648-026-10133-8