Artificial intelligence (AI) is an important paradigm which is having a powerful impact on many fields, including education. There is a lot of confusion and misunderstanding around the term, and so the purpose of this article is to provide some concrete examples of how AI is being used to improve the language learning experience, and why it is something for language teachers to embrace rather than fear.
What AI is not
Most articles about AI come with an obligatory picture of a humanoid robot, setting up the idea that artificial intelligence is all about recreating the human brain inside a machine. The truth is a lot more prosaic. Most artificial intelligence applications have nothing to do with robots or replicating human cognition; instead they focus on using the vast computational capabilities of modern computers to solve single, simple problems in a much more effective way than a human can. Although human brains are capable of doing many very complicated things in much more sophisticated ways than computers, they are limited by factors such as the size of our working memory or the speed at which we can make calculations. For some tasks, especially ones involving many or complex calculations, computers are much more effective and this is where AI steps into play.
What AI is
Artificial intelligence crosses many different domains, including fields such as computer vision (think self-driving cars which can distinguish a pedestrian from a signpost) and predictive analytics (think about the Facebook timeline, which predicts what story you want to see next based on what you’ve clicked on before). In fact, a more helpful term for thinking about what artificial intelligence can do is machine learning. Machine learning is when humans (generally data scientists) use large data sets to train computers to make models which predict the outcome of some future event. This could be the likelihood of it being safe for a self-driving car to make a left turn, or – to use a language learning example – the likelihood that a student will know how to translate the word chien into English.
When you look at machine learning in more detail, it is fundamentally statistics. When we talk about AI, we are often referring to the power of simple, well-established mathematical formulas to make accurate predictions. These techniques have been around for centuries in some cases; what has changed is our access to the computing power to automate calculations and use them in innovative new ways such as self-driving cars, personalised content recommendations or adaptive learning systems.
How is AI helping language learners?
AI has the potential to make digital language learning truly personalised to each learner: reducing the time, cost and frustration involved in completing online or app-based courses. We have recognised this at busuu for some time, and have been investing in our data science team as well as upskilling our team of linguists in the potential for tools like machine learning.
As an initial step, we recently relaunched our Vocabulary Trainer feature with a unique machine learning algorithm which adapts in real time to learner behaviour, calculating each learner’s vocabulary strengths and weaknesses and generating an entirely personalised set of study materials in each session. Importantly, the algorithm learns from both individual and collective learner behaviour, making its predictive power very strong. For example, our model learns from user data that most students have no problem translating the word ‘merci’ on their first attempt, but that they take an average of three attempts before they can correctly translate the word ‘lunette’. These data allow us to build a collective probability score for each word, and to predict how many times the average learner needs to be tested on that word, and how long to wait between testing sessions. Once we start to collect data on each individual learner and how they actually perform during a Vocabulary Trainer session, we can then adjust the model to take into account their individual strengths and weaknesses.
All this might sound over-complicated, but what it essentially allows us to do is save the learner time, by only testing them on words they are likely to get wrong, rather than words that they are likely to get right. If you’ve ever used an online vocabulary testing application and become frustrated at being tested on words that you’ve already committed to memory, you’ll recognise the potential of this technique!
What’s next for AI in language learning?
Based on our success with vocabulary, we are collecting learner data on other language areas, such as grammar. These data are being used to train additional machine learning models that will allow us to create personalised practice and review sessions for different skills, giving learners even more targeted information on their areas of strength and weakness.
We foresee a future where one-size fits all language training solutions effectively disappear, and language learning apps adapt in real time to learners as their skills develop, making the experience of learning online completely tailored for each learner. In support of this future, the entire busuu team has taken part in training to understand and implement the potential of AI and machine learning. Our team of data scientists, product designers and linguists work closely together to benefit from each others’ expertise and to design courses and features which have AI built in from the first prototype.
How can AI help language teachers?
First of all, AI is a fascinating topic, and one which can provide rich discussion in conversation classes. There are fantastic resources online for explaining what AI really is if you want to discuss the technical implementation of tools like machine learning. Alternatively, there are some intriguing visions of dystopian futures or concepts such as the singularity, that make for lively debate topics.
Most importantly, the future of AI is about saving both learners and teachers time to focus on developing skills such as conversational fluency or confidence in communicating across cultures. The strength of AI is in creating a personalised learning experience that allows the student to work on their personal areas of weakness and benefit from tailored feedback, rather than following along to a one-size-fits-all model of learning. If students are using AI-powered language learning tools in their own time, their classroom time can be optimised for focusing on the skills and capabilities that no machine can (yet!) deliver for them.
If you’d like to learn more about AI without getting too deep into the technical details, this great free course from the University of Helsinki provides a nice introduction to the topic, and is suitable for anyone, regardless of prior knowledge.
Designed by linguists, busuu combines human interaction and AI-powered features to help you learn a language faster. There are 12 languages available to more than 90 million learners worldwide.
Interested in ten hours of personalised tutor support, focused on AI for language learning or any other area of ESL? Check out our Teacher Coaching Program here.