Integrating AI-Assisted Simulations to Enhance Economics Education Outcomes

Thesis Abstract

This study investigates how AI-assisted simulations can improve learning outcomes in undergraduate economics education by addressing gaps in experiential understanding of micro- and macroeconomic concepts and decision-making under uncertainty. The problem arises from persistent gaps between theoretical instruction and real-world economic reasoning, as evidenced by declining student engagement, low problem-solving transfer to novel contexts, and suboptimal performance on applied econometrics tasks. The aim is to evaluate whether integrative AI-driven simulations—featuring interactive agent-based markets, dynamic policy environments, and real-time feedback—enhance conceptual mastery, analytical skills, and motivational engagement compared with traditional lectures and static virtual labs. The specific objectives are (1) to determine the impact of AI-assisted simulations on course-embedded learning gains in microeconomics and macroeconomics; (2) to assess improvements in students’ decision-making accuracy under uncertainty using simulation-generated data; (3) to examine changes in higher-order cognitive skills such as hypothesis formulation, data interpretation, and policy evaluation via integrated analytics dashboards; (4) to explore student perceptions of instructional authenticity, self-efficacy, and perceived usefulness of AI-based tools; and (5) to identify contextual factors moderating the effectiveness of simulations, including prior quantitative background and instructional design features. A mixed-methods design is employed, combining quasi-experimental and interpretive approaches. The population comprises undergraduate economics majors enrolled in two large public universities over two consecutive semesters. A total of 320 students will be recruited, with 160 in the experimental group using AI-assisted simulations and 160 in the control group receiving conventional instruction augmented with static online simulations. Data collection instruments include validated achievement tests covering micro- and macroeconomic concepts (pretest and posttest), standardised course assessments, and simulation analytics capturing decision quality, time-on-task, and policy evaluation accuracy. Additional instruments include a Likert-scale survey measuring motivation, self-efficacy, perceived realism, and perceived usefulness of AI tools, as well as semi-structured interviews with a purposive subsample of 24 students and 6 instructors. Data collection will occur at three points baseline, mid-semester, and at course end. The primary quantitative analyses will use ANCOVA to compare posttest scores while controlling for pretest performance, multivariate regression to examine drivers of learning gains, and repeated-measures ANOVA to assess trajectory effects. The study will also employ hierarchical linear modeling to account for nested data (students within class sections). Qualitative data will be analyzed using thematic analysis guided by grounded theory principles to identify emergent themes about authenticity, cognitive engagement, and perceived affordances of AI simulations. Expected findings include significant improvements in learning gains for the experimental group, particularly in applied problem solving, data interpretation, and policy evaluation tasks, with effect sizes in the medium to large range (Cohen’s d ? 0.50–0.75). Simulation-driven decision accuracy under uncertainty is anticipated to outperform traditional methods, and engagement metrics are expected to show higher motivation and self-efficacy. Qualitative insights are likely to reveal that realism, adaptability, and immediate feedback foster deeper conceptual understanding and greater perceived relevance of economics content. The study contributes to knowledge by providing causal evidence on the effectiveness of AI-driven pedagogies in economics education, clarifying how AI-assisted simulations mediate learning through authentic problem contexts, feedback loops, and data-informed reasoning. It also expands the theoretical discourse on technology-enhanced learning in economics by integrating concepts from constructivist theory, situated cognition, and the Technology Acceptance Model (TAM) with agent-based simulation frameworks. Based on findings, practical recommendations include scalable design guidelines for integrating AI-assisted simulations into introductory and intermediate economics curricula, refinement of assessment rubrics to capture higher-order competencies, and professional development for instructors in AI-enabled pedagogy. The study concludes that well-designed AI-assisted simulations can significantly enhance conceptual understanding and applied economic reasoning when embedded within coherent instructional sequences, with the most pronounced gains observed among students with moderate prior quantitative preparation. Limitations include potential variability in instructor adoption, technological infrastructure constraints, and the need for ongoing calibration of simulation complexity to align with course outcomes. Future research should investigate long-term retention, cross-disciplinary transfer of simulation-based skills, and the differential impact of AI personalization levels on diverse student populations.