What the Beijing Consensus on Artificial Intelligence and Education gets right, and where it misses the mark

  • September 26, 2019

We have been rifling through the pages of the Beijing Consensus on Artificial Intelligence and Education, which emerged from a UNESCO conference back in May. It is hard to fault the itemised declarations in the document: it makes all the right noises about ensuring that AI in education should be “human-controlled”, that “human interaction and collaboration between teachers and learners must remain at the core of education”, that AI tools should support…”adaptive learning processes”, “leverage the potential of data” and “support large-scale and remote assessment” and that “technological breakthroughs in the field of AI in education are an opportunity to improve access to education for the most vulnerable groups”. It even recognizes that “AI should be designed in an ethical, non-discriminatory, equitable, transparent and auditable manner”; and that AI applications in education should be developed that are “free from gender bias”… and use “gender sensitive data”.

That said, we would note that the conference was claimed to have brought together over 500 international representatives from more than 120 Member States, including 50 government ministers and almost 100 representatives from UN agencies, academic institutions, civil society and the private sector. An impressive gathering indeed. However, on detailed examination of the list of speakers and panellists, a number of these constituencies were narrowly based, for example, most of the Ministers were from Africa, China and its near neighbours, almost all of the EdTech suppliers were Chinese and only Microsoft was present from the big US tech companies. While there was widespread representation from universities round the world including Assistant Professor in Learning Sciences and Innovation at The Open University, Dr Wayne Holmes (read our interview with Dr Holmes here), not a single school teacher from any country presented or participated in the panel discussions. Nor did the conference programme seem to include any discussion of existing programmes, anywhere in the world, in which AI is already being used in search of improved educational outcomes.

Perhaps this is because existing AI programmes are seen as too limited to simply personalising student pathways towards the same old exams targeted by today’s worldwide schooling model when AI should in time allow students and teachers to collaborate on a whole new educational destination. All the more reason to have heard teachers express their views on how they believe technology could best assist them to unearth the buried treasure in their students.

So while the wording of the so called Beijing consensus was unobjectionable, we feel it was of limited value since the conference organisers missed an opportunity to invite teachers and AI suppliers with real experience of using AI in education, and so did not address the practicalities of more widespread adoption of AI in education, let alone of its more revolutionary possibilities.