Beyond Adoption Rates: Sustainable AI-Integration in K-12 Education

Thursday, July 16, 2026, 7:00 PM

Session: SOC Plenary Virtual 2 (Virtual)

Teacher AI adoption is growing but remains fragile. Once adopted, AI tools may be abandoned, leading to competency loss through forgetting and requiring a second round of professional development before re-adoption is viable. This dynamic is invisible to cross-sectional survey methods and unaddressed by existing technology adoption frameworks. To the best of the authors’ knowledge, this paper develops the first system dynamics model of teacher AI adoption calibrated against TALIS 2024 data and other secondary sources. The model uses a co-flow structure to track teacher competency across three adoption states over a 100-month simulation horizon. Eight policy experiments are tested. Policies that accelerate adoption, including peer network amplification and institutional mandates, reliably generate a competency trap in which peak disengaged teachers reach nearly double the baseline level. Training without embedded classroom practice produces worse outcomes than the status quo across every indicator. Re-engagement support for previously disengaged teachers is the most efficient single-lever intervention available. Disengaged teachers carry substantially higher workloads than their adopting colleagues, creating a growing within-school workload inequality. Sensitivity analysis confirms that these findings are robust across the plausible parameter space. The findings have direct implications for how AI adoption strategies are designed, sequenced, and evaluated in national education systems.