AI-assisted formative assessment for economics education equity

Thesis Abstract

The rapid integration of artificial intelligence (AI) in educational practice has raised concerns about equity in economics learning, particularly regarding formative assessment that adaptively supports diverse learner needs across socioeconomic and demographic groups. This study addresses the problem that AI-enabled formative assessment tools may reproduce or exacerbate existing inequities in economics education by uneven access, differential feedback quality, and biased performance metrics. The aim is to evaluate whether an AI-assisted formative assessment system can enhance learning equity in introductory economics courses and identify the mechanisms through which it operates. Specific objectives are (1) to design and deploy an AI-driven formative assessment platform that provides adaptive quizzes, instant feedback, and individualized learning resources; (2) to examine differences in learning gains, engagement, and perceived fairness across students from differentiated backgrounds; (3) to investigate the relationship between feedback specificity, cognitive load, and equity outcomes; (4) to explore instructors’ and students’ experiences with the platform and perceptions of transparency and trust; and (5) to develop a theory-driven framework linking AI-assisted formative assessment with equity in economics education. A mixed-methods research design combines a quasi-experimental approach with qualitative inquiry. The population comprises first-year undergraduate economics students enrolled in three large public universities. A total sample of 600 students will be recruited, with 300 assigned to the AI-assisted formative assessment condition and 300 to a traditional formative assessment control condition, matched on prior achievement, gender, socioeconomic status, and ethnicity. Data collection instruments include (i) standardized economics achievement tests administered at baseline and after a 12-week intervention, (ii) learning analytics from the AI platform capturing time-on-task, hint usage, and resource access, (iii) validated scales measuring perceived fairness, self-efficacy, and motivation, and (iv) semi-structured interviews and focus groups with a purposive subset of 40 students and 12 instructors. Validity and reliability will be ensured through pilot testing, Cronbach’s alpha for survey instruments, and triangulation across data sources. Quantitative data will be analyzed using hierarchical linear modeling (HLM) to account for nested data (students within classes) and multiple regression to test mediation and moderation effects. ANCOVA will be employed to compare post-test performance while controlling for baseline scores. AI-generated feedback features will be analyzed for equivalence in quality across student subgroups using content analysis of feedback logs. Qualitative data will be analyzed through thematic analysis guided by the expectancy-value and social cognitive theory frameworks, with coding performed by two researchers and cross-verified for intercoder reliability (Cohen’s kappa > 0.80). Expected findings include (i) higher learning gains and reduced achievement gaps in the AI-assisted condition, particularly for students from historically underserved groups; (ii) greater engagement and perceived fairness associated with immediate, tailored feedback and transparent explanations of problem-solving steps; (iii) evidence that feedback specificity and cognitive load balance mediate equity-enhancing effects; and (iv) positive perceptions of trust and explainability of AI feedback among students and instructors, contributing to sustained use. The study contributes to knowledge by providing empirical evidence on the equity implications of AI-assisted formative assessment in economics education, clarifying mechanisms linking adaptive feedback to learning equity, and informing design principles for transparent AI systems in higher education. It advances theoretical integration of expectancy-value theory and social cognitive theory within an AI-augmented assessment context and offers a practical framework for scalable implementation in diverse tertiary settings. Practical recommendations include guidelines for equitable AI design, considerations for data governance and bias mitigation, instructor professional development for interpreting AI feedback, and policy implications for leveraging technology to close equity gaps in economics education. The main conclusion is that carefully designed AI-assisted formative assessment can enhance learning equity without sacrificing instructional quality, provided that transparency, bias mitigation, and rigorous evaluation are integral to system development and deployment.