System of Least Prompts: a four-step graduated scaffolding hierarchy ending in graceful exit with direct instruction¶

Strategy · All strategies
Evidence · no claims cited

Description¶

When a student struggles, the AI proctor employs the System of Least Prompts, "offering a strict hierarchy of scaffolding": a general nudge, a specific cue, a targeted scaffold, and finally direct instruction. Exhausting the hierarchy triggers a Graceful Exit that finalizes the score for that interaction (awarding zero earned points, logged via the shadow ledger) but supplies the missing knowledge so the conversation proceeds, preventing a momentary recall lapse from capping the student's overall demonstration of mastery.

Design Implications¶

Context¶

Requirements¶

• An AI proctor that logs hint counts deterministically and can supply missing foundational knowledge without terminating the topic

Constraints¶

• Exhausting the hierarchy finalizes only that interaction's score; the proctor continues generating parallel prompts within the tier until the evidence buffer is satisfied

Target Learners¶

• higher education students being assessed conversationally

Target Learning Goals¶

• moving from unassisted to assisted performance within the zone of proximal development

Affordances¶

• Socratic Test Framework Bloom Solo Zpd

Related Strategies¶

Examples¶

-

Key Sources¶

• Ilya Mikhelson. (2026). The Theoretical Foundation of Socratic Tests: Dynamic, Multimodal, Conversational Examinations. https://arxiv.org/abs/2607.29624