AI & Digital Education

Educational systems should protect learner agency and privacy while measuring whether people understand and can apply the material. Interface activity alone is not learning.

Principais conclusões

• Define the learning objective.
• Evaluate accessibility and learner agency.
• Measure learning and privacy outcomes.

Start with a learning objective and the teacher or learner decision the system supports. A generated explanation should be accurate, level-appropriate, and open to questioning. Avoid hiding the reasoning or answer when the activity is intended to build independent problem solving. Evaluate across languages, devices, accessibility tools, and prior knowledge. A translation or automated feedback can sound fluent while changing the concept or discouraging a learner. Keep a teacher review path for high-impact assessment and show the evidence behind an explanation. Protect student work and identity. Document collection, retention, provider access, and age-appropriate safeguards. UNESCO’s guidance emphasizes a human-centered approach, privacy, and ethical validation for generative AI in education and research. Measure learning gains, completion, confidence, and support burden together. Iterate from actual learner errors and teacher observations, and keep a useful non-AI route for people who cannot or should not use the tool.

1. Imagine a tutor that immediately provides every answer, increasing session completion.
2. Measure whether learners can solve a similar problem independently afterward.
3. Change the interaction to offer graduated hints and review the learning outcome.

The constructed example distinguishes completion from understanding.