Generative AI in Education: Real Opportunities and Risks to Manage
Generative AI tools, which can produce text, images and code, have become part of the lives of many learners and teachers in a short time. They raise fundamental questions: how can we use them for learning without letting them replace thinking? And how do we protect privacy and academic integrity? This article summarises what leading international bodies and published research say, and suggests practical habits.
How do these tools work, in brief?
Large language models are trained statistically on huge amounts of text and learn to predict the next word in a given context. This is why they produce fluent, convincing text, but they do not "know" the truth in the human sense, and they can produce wrong information or non-existent references with complete confidence. Understanding this probabilistic nature is the basis of sensible use.
Opportunities worth using
In their 2023 review, Kasneci and colleagues identify a range of promising educational uses, including:
- Generating varied examples and exercises for practice.
- Explaining the same concept in different ways and at different levels to suit the learner's background.
- Giving initial feedback on written drafts before a person reviews them.
- Supporting learners in several languages and improving access for people with different needs.
- Helping teachers prepare materials and handle routine marking tasks.
Risks that must be managed
The same sources identify risks that should not be ignored:
- Errors and hallucination: producing invented information or references that look correct.
- Bias: reproducing biases present in the training data, including bias against less represented languages and cultures.
- Over-reliance: a decline in thinking and writing skills if the whole task is handed to the tool.
- Privacy: entering personal or sensitive data into systems that may store or use it.
- Academic integrity: presenting generated text as one's own work without disclosure.
What does UNESCO recommend?
In 2023, UNESCO published guidance on generative AI in education and research that stresses a human-centred approach. Its main recommendations include protecting data privacy, setting a minimum age for independent use of these tools, training teachers, and checking that tools are pedagogically and ethically appropriate before educational institutions adopt them. OECD reports on digital education also discuss how to integrate these technologies within clear education policies.
Practical principles for learners
- Use the tool as a thinking partner, not a substitute for thinking: ask it for questions that test your understanding, not ready-made answers.
- Check every important piece of information and every reference in its original source.
- Do not enter personal or confidential data.
- Disclose your use of the tool according to the rules of the body you submit your work to.
- Keep your own drafts to show how your thinking developed.
Summary
Generative AI is a powerful tool that can widen learning opportunities, provided that people remain the decision-makers and remain responsible. FUTURIA's Lab for AI and the Human Future offers practical activities for testing these principles.
References
- UNESCO (2023). Guidance for Generative AI in Education and Research. Paris: UNESCO.
- Kasneci, E., et al. (2023). ChatGPT for good? On opportunities and challenges of large language models for education. Learning and Individual Differences, 103, 102274.
- OECD (2023). OECD Digital Education Outlook 2023: Towards an Effective Digital Education Ecosystem. Paris: OECD Publishing.
- Holmes, W., Bialik, M., & Fadel, C. (2019). Artificial Intelligence in Education: Promises and Implications for Teaching and Learning. Boston: Center for Curriculum Redesign.
