In the final part of our series of interviews with Knewton, we ask about the single most common objection we hear when discussing big data and adaptive learning: can it really work for language learning, a subject that can’t simply be broken down into ‘atoms’ of knowledge that have to be acquired? Sally Searby (Partnership Manager) and David Liu (Chief Operating Officer) give us their thoughts.
eltjam:
Do you think there are any particular challenges or difficulties specifically in the subject of language learning, in terms of applying what you do? Because there is a view in ELT that this is all very well for maths or science, but language learning is different. What’s your take on that?
Sally:
With language learning, what are the prerequisites if you’re at a B1 level? You have a good grasp of language, but there’s a huge amount that you don’t have. So, what are the prerequisites to that? I think it can be done, but they’re much more nuanced. You have to take something like grammar, and say, OK, if I’m going to learn this functional language, I need to understand this grammar in order to achieve that.
What’s going to be very interesting with language learning in ELT is, the more data we get, the better we’ll understand it. But, to start with, you have to scope it. In terms of building ELT products powered by Knewton, building it from scratch is going to be more effective than trying to apply it to courses and backlist. But that’s what people have to do in the shorter term, and to begin to understand what Knewton can do and that’s the approach that I think some of the publishers are taking.
that is much more complex than many other subjects that have so far been powered by Knewton
To be incredibly effective, we’ll have to create knowledge graphs and develop a deeper understanding of what is required when learning a language. You may have a student who has a very strong grasp of reading comprehension, has a very good functional vocabulary, but their grammar may not be as strong, or they’re not so good at the writing and the productive skills. So somehow we have to grasp that and that is much more complex than many other subjects that have so far been powered by Knewton. I do think it’s very possible, but it’s going to be complex.
David:
I would agree. There are these basic areas of language learning and our premise is that we try to break everything down, and really analyse the domain at its basic level. You need to identify whether there is a rubric that has been established or can be established in terms of understanding one’s proficiency. If there’s a rubric that can be agreed to by the content developer, we strive to, and most of the time, can, make it adaptive.
In language learning, it is a bit more nuanced, but it is still based upon rubrics and understanding certain components of good grammar. And so it is not as straightforward say, as math, certainly, which is a bit more formulaic.
But we still work very effectively in subject areas beyond math, and in fact we have powered MyReadingLab with Pearson. In MyWritingLab, students learn sentence structure – you have to start with a beginning sentence and then there’s a supposition, and then you make your conclusions. Even in writing, there’s a rubric around ‘what is a good paragraph?’
if you’re building a product built for digital from the ground up, it allows you far more flexibility
We’re learning a lot right now, and I would absolutely back Sally 100%. You know, we work both with existing products and new products. We do mild changes to retrofit our technology into those existing products. But obviously there’s more that we can do when we start with a product from scratch. And we know that we don’t have the luxury to choose. The partners come with what they have, and we work equally well in both, but of course if you can start from scratch, if you’re building a product built for digital from the ground up, it allows you far more flexibility.
Sally:
The other thing which we haven’t talked so much about is the goals – or outcomes – and building that in, and the assessment, which is the direction that ELT is going in very fast with integrated assessment and instruction. And I think the more it’s based around the goals or the outcome for the students, and the Common European Framework, the ‘can do’ statements, the high stakes exams, the easier it will be to create the content and graph the content. And I think that’s a good starting place for ELT publishers, because that does limit the scope, and that does help, and a lot of work has already been done around that. I think the more that ELT publishers can use the tools that they have, like the corpus and the research that they’ve already done into language learning, the better the product will be. And I think that those kinds of tools will help in terms of graphing and creating the content for this.
the more it’s based around the goals or the outcome for the students, and the Common European Framework, the ‘can do’ statements, the high stakes exams, the easier it will be to create the content and graph the content
eltjam:
It sounds like we really just need some ELT publishers to build some products designed around this, and then we will actually start to refine it, and learn some of the things that need to happen to make it work really well for language learning. Would that be reasonable?
David:
Yeah, there’s nothing like learning from practical examples. And so I think what I’m encouraged with, so far, is beyond our existing partners in the ELT space, Macmillan and Cambridge University Press, we’re speaking with large companies in the ELT space. They are outside of Europe, in South America, Brazil, in Asia, in the US. I think that now we need to learn from one another and of course build products – and our role in this, besides being this data layer of infrastructure, that people leverage to build great products – our role is also to provide guidance. We do consult, we provide a lot of industry expertise in terms of how you can get to better outcomes, how you can leverage assessments.
And we do know instructional design, even though we do not author it. We don’t tell the publishers what to do, but we’re familiar with it. We hope we can help them make products better, whatever it is that they are doing. So, it’s actually a very strong partnership in that way.
eltjam:
Any final thoughts?
David:
We’re open for business, and no matter what ELT publisher and country you’re in, we’d love to speak with you. We have now a corpus of knowledge and experience in working with some of of existing partners that we can apply.
I think the future’s really bright for this industry, especially as adaptive learning and data can both help publishers create more effective learning materials and improve student outcomes.
Knewton interviews
Part 1 – Big data and adaptive learning in ELT
Part 2 – Sharing data and competitive advantage
Really interesting, thanks Laurie. Being a data junky, I love the theory, but I can’t help but be a little sceptical about how successful it will be in practice – it all got a bit woolly when you probed about fitting the platform to language learning. I think a system that could successfully predict learning and present relevant resources is the Holy Grail, but I’m not sure if such a rubric could exist that can model the infinite ways different learners approach language learning. One to watch though for sure!
Yes, very interesting. A major issue here is that learners – and their teachers, the publishers’ direct clients, may have many different types of desired outcomes, many of which are outside Knewton’s traditional comfort zone of high stakes learning and clear goalposts like the CEF. What is clear, though, is that instructional design should benefit enormously from intelligent, nuanced use of Knewton’s analytical tools.