AI in Education Raises Questions We Are Not Equipped to Answer

Why are current ethical frameworks inadequate for AI in education?

Existing AI ethics frameworks were designed for decision systems (approve/deny), not for formative systems (shape how a person develops), and education AI is fundamentally formative in ways that existing frameworks cannot address.

I evaluated 4 AI tutoring platforms in 2025. Each had a privacy policy. Each had a fairness statement. None had a coherent answer to the question: “What theory of learning does your system encode, and who decided that theory was correct?” Every adaptive learning algorithm embeds assumptions about how learning works. Those assumptions shape what content a student sees, when they see it, and how their performance is evaluated. These are pedagogical choices with profound implications, made by engineers, not educators.

A lending algorithm that denies a loan application produces a discrete outcome that can be audited. An AI tutoring system that shapes how a 12-year-old understands mathematics produces diffuse, long-term effects that may not be observable for years. Our ethical frameworks are built for the first type of system. They are not equipped for the second.

What specific ethical questions does AI in education raise?

AI in education raises 4 categories of ethical questions that existing frameworks fail to address: pedagogical authority, surveillance of minors, cognitive formation, and equity of access.

• Pedagogical authority: Who has the right to decide how a child learns? When an AI system selects content, adjusts difficulty, and evaluates understanding, it exercises pedagogical authority. I found that none of the 4 platforms I evaluated had formal educator involvement in their adaptive algorithm design. The learning path was determined by engagement metrics, not educational theory.
• Surveillance of minors: AI tutoring systems collect granular data on how students think: time spent on each problem, error patterns, attention indicators, and emotional state (in systems with webcam access). This creates detailed cognitive profiles of minors. The FERPA framework was not designed for this level of surveillance.
• Cognitive formation: An AI system that consistently presents mathematics as procedural problem-solving rather than abstract reasoning shapes how students think about mathematics. The ethical question is not whether the system is accurate but whether we have the right to make these formative choices algorithmically, and who bears responsibility for the outcomes.
• Equity of access: AI tutoring systems perform differently based on available hardware, internet connectivity, and the student’s language proficiency. Students in under-resourced environments receive a degraded version of the tool. This is not just a technical limitation. It is an equity problem that compounds existing educational disparities.

How should the AI ethics community approach these questions?

The AI ethics community should develop education-specific frameworks that center the developmental needs of students, require meaningful educator participation in system design, and establish surveillance boundaries for minors.

I do not have answers to these questions. I have observations from implementing AI systems in educational contexts, and those observations tell me that the current approach is insufficient. The evaluation methods we use for AI systems are designed to measure accuracy and fairness. They are not designed to measure developmental impact, pedagogical appropriateness, or the long-term effects of algorithmic learning paths on cognitive development.

According to UNESCO’s guidance on AI in education, “the use of AI in education raises unique ethical considerations related to the rights and developmental needs of children.” The guidance is correct. The implementation is absent. We are deploying AI tutoring systems to 120 million students while the ethical frameworks that should govern their design remain in draft form.

What is at stake if we continue deploying without adequate ethical frameworks?

The risk is not a single ethical violation but a generation of students whose educational experience was shaped by algorithmic systems that no one fully evaluated for developmental appropriateness.

This is not an argument against AI in education. It is an argument for intellectual honesty about the gap between our deployment speed and our ethical readiness. The questions I have outlined do not have easy answers. Some may not have answers at all, only tradeoffs to be navigated with care. But navigating tradeoffs requires first acknowledging them, and the current discourse around AI in education treats ethical complexity as a problem to be solved rather than a condition to be managed.

I believe AI can improve education. I also believe we owe students, particularly young students, a more rigorous ethical foundation than what currently exists. The honest position is that we are not equipped to answer the questions this technology raises. The responsible action is to deploy more slowly, evaluate more carefully, and invest in the interdisciplinary frameworks that education AI demands.