New this week from IELTS is a short article which explores “the potential of conversational AI in speaking assessment.”
Nothing really earth-shattering here, as it mostly summarizes longer stuff they’ve put out over the past little while. I liked this part at the end, though:
“…traditional rule-based SDSs, as opposed to their AI-powered counterparts, therefore seem to be a more practical choice in the short term, given their greater consistency, scalability, and control over test delivery.”
It is mildly interesting that IELTS can only bring itself to commit to traditional human interlocutors in “the short term.” One of these days, I think, they’ll announce a big change. Don’t say that Old Man Goodine didn’t warn you.
A silly thought experiment might be to estimate how much it costs to maintain the current method of speaking assessment. It certainly isn’t cheap – the global average is probably close to $12 USD in examiner wages per test. On top of that are the costs of payroll taxes, training, recruitment, management and physical infrastructure. The proposed AI model would still involve ratings by humans, but offshoring that process could bring the cost down to just a few bucks per test. Or less. The partners would love it – I can’t help but think such a move would be just as lucrative for IDP Education as, say, opening a bunch of new test centers in China.