4 Comments
Dec 6, 2023Liked by Johan Fourie

Another week, another thought-provoking masterpiece. Lots to unpack here. Will definitely have to return to this article a few more times and share it with the rest of my network. 🙏

Expand full comment
author

Thanks for the kind words, Taf! Hope you are doing well.

Expand full comment
Dec 4, 2023·edited Dec 4, 2023Liked by Johan Fourie

The long history of failed Ed-tech interventions should perhaps dissuade one from too much hope in AI, particularly for foundational literacy and mathematics. Early learning is about developing a socio-emotional relationship with a teacher. No matter the emotional intelligence of a computer screen, I don't think any 5-year-old is going to believe they can excite or let down their AI tutor behind a screen. The skill of sitting on your bum and listening for many hours is completely avoidable by walking out of the room, should there not be a teacher with whom you've formed a relationship. These sorts of skills are what one is really being taught in early grades.

My most effective maths teacher wasn't the teacher who had the best pedagogical knowledge, but the one I liked the most (see Hanushek's work on educator unobservables). For more evidence on this, first language acquisition is dependent almost entirely on the number of words a child hears. Yet, if those words originate from radio or TV, they count for naught (part of the "1 million words" findings). It is in the anticipation of responding, to be heard, that leads to words being learnt. The most effective Ed-tech interventions have been providing teachers with highly prescribed class plans, so at best, Ed-tech cannot rid us of the biggest cost of education (teachers, which cost 1 in 10 tax Rands).

As foundational literacy and numeracy is SA's biggest hurdle, I think our binding constraint won't be diminished by AI. There is certainly scope for university.

I am also dubious about most growth claims. But at the micro-level, returns to Ed are massive in SA. The skills deficit is a margin employers often cite as a constraint in SA. Bantu Ed was so bad, we’ve had the most Ed improvement in Africa and still haven’t caught up with Kenya. The persistence literature shows human capital scarring is particularly persistent. I think the evidence for human capital composing a large part of the Solow residual is also good, even for Vietnam. Their education may have been better than SA's, but that is an extremely low bar. There has been incredible improvement in step with growth (esp. when controlling for, e.g., USA FDI), and cross-sectional evidence on growth and change in learning is very strong.

Expand full comment
author

I hear you, Peter, and share your caution in relying exclusively on AI. But the 'long history' of failed edtech experiments pale into insignificance when compared to the long history of failed public interventions (or failure to implement public interventions, for a variety of reasons). If we simply rely on the state to address the education crisis, we will continue to have generations of kids with poor results. We need to experiment with alternatives, and given the speed at which AI is improving, I'm willing to take that bet.

Expand full comment