#164: Simulations to Facilitate AI Literacy for Teachers and School Leaders

We developed a novel web-based simulation targeting the concept of feed-forward neural networks for an adult educator audience comprising two distinct cohorts of teachers (N = 24) and school leaders (N = 23). Both cohorts showed gains in their conceptual knowledge from pretest to posttest, with the school leaders cohort improving relatively more (Cohen’s d = 0.40). Comparative analyses suggested that using our simulation led to lower perceptions of usability and learning effectiveness in school leaders relative to teachers, but higher actual as well as normalized learning gains on the administered performance assessments. These results highlight a perception-performance discrepancy when learners grapple with less-than-perfectly-intuitive interfaces, which despite resulting in low fluency, lead to better learning.