Duolingo Versus One-on-One Tutoring

Luis von Ahn, CEO and Co-founder of Duolingo

I recently found a video in my Youtube subscription feed that I thought, from the title and description, would be brilliant. The video was from Big Think, and it was titled “The Future of One-on-One Education, with Duolingo’s Luis von Ahn”. I know a great deal about von Ahn, and about Duolingo, as I researched them both during my technology in education studies. At the time, Duolingo had just been announced but not yet launched, and von Ahn was mostly known for his invention of the Captcha system for distinguishing actual people from machines and algorithms.

The idea for Duolingo is strongly linked to his Captcha project. In the TEDx talk at Carnegie Mellon University where he announced the service, he talked about how he felt bad about having wasted “500,000 hours every day” of humanity’s time by making people solve 200 million captchas per day. So with the next version of the Captcha system, later dubbed re-Captcha, he set about to solve that. And the solution was ingenious: With re-Captcha, the solutions that people give help computers digitise texts by, essentially, putting unresolvable snippets of text up to a “vote”. A re-Captcha has two parts; one word picture that is known, and one that is not. The user doesn’t know which is which, so they have to solve both. Get the known one right, and you’re human – and by guessing on the unknown one, you’re helping computers interpret it.

Duolingo works along similar lines of reasoning. With it, von Ahn wants to “translate the web”. Doing this in the traditional way – by hiring professional language translators – would be prohibitively expensive. But von Ahn realised that there are millions, probably billions, of people out there who are actively trying to learn a second (or third etc.) language. He cites a figure of 1.2 billion people. Many of them are willing to spend a lot of money on software to help them achieve their goals, with over 5 million Americans having paid over $500 USD for common software solutions like Rosetta Stone. With Duolingo, von Ahn believes he has found a better way. The idea is that the translation parts – the “work” that they want the translators to do – is woven into a learning experience. With Duolingo, the users learn a language and the company gets paid by having the learners perform translation work for them.

From a language learning perspective, Duolingo is an excellent tool. It is heavily gamified, which I like very much, and it has been shown to be very effective. As is mentioned in the Big Think video, there are studies that show Duolingo is “as good as a classroom”. This isn’t particularly surprising, for several reasons: First, because classroom teaching tends toward ‘teaching to the middle’, with the strongest and the weakest students left behind. This is a natural consequence of trying to fit X students to a single set of lessons. It is also why all good language teachers will tell you that the most important work you can do to improve is done outside the classroom. This, in my opinion, is where Duolingo shows its value. However, Luis von Ahn wants it to be more than that.

It is well established that one-on-one tutoring is dramatically superior to classroom learning. The famous 2-Sigma studies he mentions early in the video are real, and clear proof of this. It also just makes sense, rationally: One-on-one tutoring gets rid of the teach-to-the-middle problem, because the single student is the middle. In von Ahn’s vision, Duolingo solves the scalability problem with one-on-one tutoring by making the service as good as the human-to-human variant. This is a great idea, and a good goal to have. There are a few problems with the idea, though, and the biggest problem with it, is that it’s never going to happen. Let me explain why.

Machines can do a lot of things, and they will certainly take over a lot of our jobs even in the relatively short term. However, they cannot replace the human element – and that is something that language learning requires. Language exists solely to facilitate communication between interpreting entitites, normally between humans. The properties of the language used changes based on the communicating participants. There are too many factors to count: Shared language basis, culture, gender, geography, relative age, context… the list goes on. And this is why human language teachers have an advantage that no computer is ever likely to have (until they, too, become “human”): They adapt to these factors instinctively and continuously in the same way as the student does, and can adapt their teaching to fit.

There is great learning potential in interaction, particularly in language learning. As a language teacher, I can tell you that it is exceptionally easy to spot the difference between a learner who interacts frequently in the L2 (the target language) and one who does not. Even if they are otherwise of similar levels, even if the one who does not use the L2 much has otherwise excellent learning strategies, the one that interacts is likely to have a much better progression than the other. And this is what makes the very idea of software replacing the human interaction element in language learning a no-go.

I am a big proponent of embracing technology and technological solutions in most teaching environments. Language learning is not an exception. However, I also feel it is both dangerous and counter-productive to think of any technology as a panacea. I have used, and will continue to use, Duolingo in my teaching and in my personal language learning endeavours. The benefits are great and many, and it is hugely effective as a language learning tool. But a replacement for human interaction it is not.