Generative AI has disrupted education. Here’s how it can be used for good – UNESCO

Barely recovered from the pandemic, schools and universities are beginning the 2023-24 academic year with a fresh challenge to contend with – generative AI.

While some educational institutions initially banned the use of ChatGPT when it launched in November last year – mainly due to concerns around how it could enable cheating – now it is a case of establishing ground rules and guidelines around generative AI use.

The first ever global Guidance for Generative AI in Education and Research, a UNESCO initiative designed to “address the disruptions” caused by the technology, is now available to help this process. As countries continue to work out their governance approach, the guidance “will help policy-makers and teachers best navigate the potential of AI for the primary interest of learners”, says Audrey Azoulay, UNESCO Director-General.

The World Economic Forum and AI Commons have similar goals with their 30 action-oriented recommendations outlined in the Presidio Recommendations on Responsible Generative AI. These guidelines – the result of 100 thought leaders and practitioners coming together at a global summit in April 2023 – are centred around three key themes: responsible development of the technology, international collaboration, and social progress, which incorporates “educational shifts”.

Fewer than 10% of schools and universities currently have formal guidance on AI, says UNESCO. Its new guidance suggests eight specific measures educational institutions could adopt to ensure “quality education, social equity and inclusion”.

Bias was identified early on as an issue with generative AI models, and the task of governments and institutions now is to level the AI playing field. To this end, UNESCO says steps must be taken to ensure everyone has internet connectivity and access to AI applications; criteria must be established to eliminate gender or cultural bias; and GenAI systems must evolve to include data in multiple languages, especially minority ones.

There is a danger we may become too dependent on generative AI and lose sight of our human agency, compromising “the development of intellectual skills”, UNESCO warns. One approach the report advises to counteract this possibility is selective banning of the technology in situations “where it would deprive learners of opportunities to develop cognitive abilities and social skills through observations of the real world”.

This point acknowledges the need for ongoing monitoring of generative AI as the technology continues to evolve. Checks and balances must be put in place to ensure “GenAI systems used in education and research are free of biases”, and that children and “vulnerable learners” understand informed consent. The report adds that “institutions and educators should be willing and able to take swift and robust action to mitigate or eliminate … inappropriate content”.

UNESCO says governments should commit to providing AI curricula in schools and colleges, “as well as for lifelong learning”. Tailored programmes should also be provided for older workers and citizens “who may need to learn new skills”.

It’s not just students who are making use of generative AI, of course. A recent US poll found that more teachers than pupils were using it, with one teacher admitting that MagicSchool – an educators’ generative AI tool – had enabled him to “take back his summer” after planning a year’s worth of lessons using the technology. Yet UNESCO says only seven countries are currently focused on AI training programmes for teachers and that future policies must “enable teachers to create specific GenAI-based tools to facilitate learning in the classroom and in their own professional development”.

Learners and educators need to approach GenAI with the understanding it is “a fast but frequently unreliable source of information” mostly trained on “dominant worldviews”. Knowing how to objectively critique the technology’s responses is vital to protect “minority opinions and plural expressions of ideas”, says the report.

Part of AI’s bias is that GenAI models are mostly trained on information from the northern hemisphere, meaning the Global South and indigenous communities are under-represented. UNESCO is calling on GenAI tools to “be made sensitive to the context and needs of local communities”, recommending that use cases in education and research accurately reflect educational priorities, “rather than novelty, myth or hype”.

AI providers, educators, researchers, as well as parent and student representatives, need to collaborate on “system-wide adjustments” across curricula to both mitigate the risks and harness the potential of generative AI, says UNESCO.

“We all need to be cognizant that GenAI might … change the established systems,” concludes the report. AI tools should not undermine, conflict with or usurp us, states UNESCO. To ensure that doesn’t happen, “the transformation of education and research [by GenAI] should be rigorously reviewed and steered by a human-centred approach”.