24/08/2023
AI is transforming English language learning and assessment, but it’s essential that teachers understand its limitations to get the very best out of it, according to an expert from Cambridge University Press & Assessment.
Speaking on the recent Teachers Talk Tech podcast Dr Jing Xu highlighted pioneering work coming out of Cambridge in AI for English language education and explained the crucial role English teachers will play in the future.
‘AI has been penetrating the English language classroom for the last ten years and offers fantastic ways of enhancing and personalising learning,’ commented Dr Xu. He continued: ‘AI has reached the level where it can accurately learn patterns in human language, which means it can be used to mark essays and for creating classroom content at the right levels for different students. It’s doing a pretty good job at this and brings a lot of value to the classroom. However, we have to be realistic that there are still limitations with AI, and that’s why teachers and examiners will still play a critical role in the future.’
Can we trust AI?
The episode, which also featured education expert Bob Godwin Jones, asked the question: can we trust AI? The speakers traced the surprisingly long history of AI, which dates back to work that was carried out in the 1940s. They also discussed the current research that’s out there, its use in English language learning and assessment and the rise of machine learning and large language models such as Chat GPT.
Cambridge’s pioneering AI
On the work that Cambridge is carrying out in this area to help teachers and learners, Dr Xu said:
‘We already use AI to generate instant feedback for English language learning. For example, our Write and Improve tool does a fantastic job of giving students instant feedback on their writing performance. This can really help students to carry out self-regulated learning and preparing for tests such as IELTS and B2 First.’
The tool was trained on the Cambridge Learner Corpus, a large collection of essays written by English language learners and annotated by human experts, using a machine learning technique called discriminative preference ranking. Simply put, the tool learns from the human annotations in the training data including marks and error identifications to predict learner proficiency and flag errors in writing.
Keep the human in the loop
However, Xu, who is ‘cautiously optimistic’ about AI in English language education, said it is essential that there is always a human in the loop to some extent. Xu commented:
‘Teachers need to be patient and use AI very cautiously. At Cambridge we are promoting an AI model, where the human judgement can step in to ensure accuracy of marking. Of course, AI brings so much value to the classroom but obviously if students receive the wrong feedback, this can be very demotivating. From this point of view teachers play a critical role in gatekeeping to ensure all learners get the very best out of AI.’
Listen to the whole episode, and read about the whole series of Teacher’s Talk Tech podcast.