Empirical Presets
Select real-world experimental cohorts or test custom memory interventions.
Cognitive Mechanics & AI Sliders
75% (Direct Solutions)
0% (Socratic Only)
50% (Draft & Polish)
100% (Direct Answers)
1 session / wk
0 days (Pure Passive)
3-4 days
7 days/wk
14 days
25% (Low Friction)
24 weeks (6 Months)
Perceived HW Grade
+18.2%
Surface performance boost
Unassisted Exam Retention
-20.4%
Independent mastery penalty
Illusion of Competence Gap
38.6 pts
Perceived vs actual recall delta
Memory Half-Life
5.8 Days
Bjork-Ebbinghaus decay pace
Longitudinal Performance Trajectory
Comparing weekly assignment grades (with AI assistance) versus unassisted retention probe score
Homework Score (With AI)
Unassisted Exam Score
Unassisted Peer Baseline (70%)
Direct Answer Delegation High Risk
Students copy prompt prompts into LLMs and submit verbatim outputs. Bypasses generation effect; neural encoding pathways remain dormant while illusion of competence peaks at +20-40% above exam truth.
Result: Rapid immediate completion, rapid decay (half-life < 4 days).
Socratic Critique Mode Balanced
AI is constrained to ask targeted prompting questions rather than supplying direct answers. Forces partial cognitive retrieval while scaffolding difficult steps.
Result: Stable homework quality (+8%), exam parity preserved (+2% to -2%).
AI-Assisted Closed-Book Recall Optimal Synergy
Students attempt problem sets unassisted first, followed by AI diagnostic analysis for blind spots and automatic spaced active-recall flashcard synthesis.
Result: +12% unassisted exam performance over traditional studying alone.
Scientific Foundation:
This simulator combines the empirical findings reported in The Economist (+18% HW gain vs -20% unassisted exam gap over 6 months) with the Bjork New Theory of Disuse (storage vs retrieval strength) and Ebbinghaus forgetting functions. Cognitive offloading eliminates the "generation effect" (Slamecka & Graf, 1978), creating high surface fluency without corresponding long-term synaptic potentiation.