The Metacognitive Outsourcing Effect: A Computational Simulation of Metacognitive and Mathematical Reasoning Development in K-12 Students Under Adaptive Platform vs. Traditional Instruction

Abstract

AI-adaptive mathematics platforms are widely deployed in K-12 public schools, yet their effects on metacognitive monitoring accuracy and mathematical reasoning transfer remain unexamined by computational modeling. This paper introduces the Metacognitive Outsourcing Effect (MOE), defined as the suppression of student metacognitive development when adaptive platforms assume the self-monitoring function on behalf of the learner. An original agent-based simulation models cognitive development for 3,000 synthetic K-12 students across five conditions -- Traditional Direct Instruction (TDI), Student-Centered Learning (SCL), Adaptive Platform Only (APO), Adaptive Platform supplementing Direct Instruction (AP+TDI), and Adaptive Platform supplementing Student-Centered Learning (AP+SCL) -- across three cohorts (Elementary K-5, Middle School 6-8, High School 9-12) over 180 simulated school days. SCL produces the strongest metacognitive outcomes (Elementary M = 0.920, d = 6.55 vs. APO); APO produces near-ceiling knowledge (M = 0.991) but the lowest metacognitive development (M = 0.317). Student-centered pedagogy substantially mitigates but does not fully neutralize the MOE when combined with adaptive platforms (AP+SCL vs. SCL: d = 1.11, 0.97, 0.68 across cohorts). A developmental gradient confirms youngest students incur the greatest metacognitive cost. Five falsifiable hypotheses are derived for empirical investigation.

Keywords