The debate about AI in education often gets stuck between two extremes: “it will change everything” or “it will ruin learning”. The realistic question is simpler: how can AI make human effort more meaningful, more accessible and more effective in solving a specific learning problem?
AI is a learning partner, not a teacher
Rather than replacing the teacher, positioning AI beside the teacher and the learner is a healthier starting point. A well-designed AI tool can patiently answer a learner's question, generate practice, give draft feedback on writing and reduce routine work for the teacher. But pedagogical decisions, learning goals and final assessment must remain in human hands.
Opportunities: from personalisation to accessibility
- Personalised learning: content and practice suggestions based on the learner's pace, prior knowledge and goal.
- Instant feedback: rapid draft-level feedback in writing, problem solving and language learning.
- Content production: fast prototyping of design materials such as scenarios, examples, questions and summaries.
- Teacher support: saving time in lesson planning, differentiation and assessment preparation.
- Accessibility: support in simplifying text, translation, captions and alternative formats.
What these opportunities have in common is that AI accelerates and personalises the learning process. But speed alone is not quality. Content produced by AI does not become a reliable learning resource until it is verified by a subject expert and an instructional designer.
Risks: the shadow of shiny automation
- Incorrect or fabricated information: language models can produce wrong content in a confident tone.
- Bias and representation: inequalities in training data can be reflected in content and recommendations.
- Data privacy: why, where and for how long learner data is stored becomes critical.
- Over-automation: unnecessarily shortening the learner's process of thinking, trying and making mistakes.
- Assessment integrity: accepting AI-generated assignments as evidence of learning.
How do you build human oversight?
Human oversight should not be thought of as an approval checkbox added at the end of the process. At which stage, within which boundaries and with which output AI will be used must be defined from the start. For example, it can be used to produce a content draft; but outcome alignment, accuracy, language and accessibility checks must remain a human responsibility.
- Define the purpose and boundaries of use in writing.
- Have every output verified by a subject expert and check the sources.
- Keep learner data to a minimum; mask sensitive data.
- Tell learners clearly when AI is being used.
- Measure learning impact and equity in the pilot.
| Use case | Role of AI | Role of humans |
|---|---|---|
| Lesson draft production | Scenario, example and practice suggestions | Outcome alignment, accuracy and pedagogical decision |
| Learner feedback | Fast draft-level responses and hints | Defining criteria, final assessment and support |
| Content accessibility | Simplification, translation, alternative formats | Context review and quality assurance |
| Learning analytics | Pattern detection and summarisation | Interpretation, ethical decisions and intervention design |
5 questions for an ethical framework
- Does this use genuinely strengthen the learner's learning, or does it only speed up the process?
- Which data is collected; who can access it and how long is it kept?
- Which biases might AI output carry; how will we test for them?
- Does the learner know they are interacting with AI?
- Which human role retains the decision and the responsibility?
Arpemit's approach
We do not put AI at the centre of education. We first define the learning problem, the goal and the evidence. AI takes its place within this framework only as a tool that strengthens human oversight and does not shorten the learner's thinking process. It brings speed to content production; but pedagogical decisions, verification and ethical responsibility remain with people.
Conclusion
AI carries genuine transformational potential in education. What will unlock that potential is not the technology itself, but the pedagogical purpose, ethical boundaries and human oversight with which we use it. Designed well, AI can become a learning partner that empowers both teacher and learner.
