Can student AI-use policies improve outcomes without widening inequality?

Can AI policies personalize learning without leaving some students behind?

The central tension is that AI can tailor instruction to each student's needs—a huge potential benefit—but the same technology can widen inequality if only some students have access to high-quality AI tools. A 2024 review found that AI can greatly improve personalization, offering activities beyond what traditional methods can provide [1]. Yet that same review warns that if only a 'specialist few' can use top-notch AI, the achievement gap could actually grow [1]. A separate global review of studies from 2013 to 2024 confirms that AI-driven solutions like personalized learning platforms have increased engagement and performance for students facing socioeconomic, geographic, and cultural barriers [3]. So the evidence agrees: AI can help, but the outcome depends entirely on whether policies ensure equal access. Without that, personalization becomes a privilege, not a right.

Can AI itself be made fair, or is bias inevitable?

A key finding is that fairness is not automatic—it must be engineered into the AI from the start. A 2022 study developed a 'fair logistic regression' model specifically to reduce bias in predicting student math outcomes on an online platform [5]. Compared to standard AI models (logistic regression, support vector machine, and random forest), the fair version achieved equally good predictive accuracy while significantly reducing errors across different demographic groups [5]. This shows that bias is not an unavoidable feature of AI; with the right design, AI can be both accurate and fair. Another 2025 study proposed a federated learning framework that predicts student performance across different regions while preserving privacy and reducing performance gaps between regions [2]. The model improved fairness by sharing insights across regions without sharing raw student data, which is a promising approach for policies that aim to close regional disparities [2]. Together, these studies demonstrate that technical solutions exist to make AI fairer—but they must be deliberately chosen and implemented.

What else must policies include to prevent widening inequality?

Even the fairest algorithm will fail if students lack devices or internet, or if teachers are not trained to use AI tools effectively. The global review emphasizes that the digital divide and inadequate teacher training are critical barriers [3]. A 2024 paper on measuring student performance with AI also stresses the need for a 'responsible and student-centered approach' and highlights ethical concerns around equity and accessibility [4]. Another review calls for policymakers, educators, and developers to work together to ensure AI is implemented following principles of transparency, accountability, and bias prevention [1]. The consistent message across all five papers is that technology alone is not enough. Policies must include investments in infrastructure, professional development for teachers, and ongoing ethical oversight. Without these, AI risks becoming another force that widens, rather than narrows, educational inequality.

About These Sources

This answer is built on 5 peer-reviewed studies — published from 2022 to 2025, 4 from 2024 or later, collectively cited 50 times — selected as the most relevant from 5 studies that passed quality screening, drawn from 49 papers retrieved from a database of over 500 million.

Sources used in this answer

Related research questions