My Duolingo learning app can reshape education
- 19 May 2014 by Aviva Rutkin
Luis von Ahn, creator of a popular language-learning app, says tracking how people learn online will reveal teaching tricks that are invisible in the classroom
Duolingo users are making new courses for people who speak Asian languages like Chinese and Hindi. How does that work?
We give people a skeleton of what the course should teach. Then they make it and we give them tools and access to a lot of data to figure out if the course is doing well and how to improve it.
We give people a skeleton of what the course should teach. Then they make it and we give them tools and access to a lot of data to figure out if the course is doing well and how to improve it.
What kinds of tools and data do you use?
For example, right now we're teaching adjectives after plurals, so let's try teaching adjectives first. We pick a subset of 50,000 users and see if they learn better – if they come back more often and make fewer mistakes. We can see the statistics and if they're positive, switch all users to the new method. We then give the community making the courses access to the same information so they can run experiments to work out how to do better.
For example, right now we're teaching adjectives after plurals, so let's try teaching adjectives first. We pick a subset of 50,000 users and see if they learn better – if they come back more often and make fewer mistakes. We can see the statistics and if they're positive, switch all users to the new method. We then give the community making the courses access to the same information so they can run experiments to work out how to do better.
What can conventional language teachers take away from these experiments?
We're in a position to discover how people learn on a much larger scale, and we're going to release a lot of that information. Let's say moving a single word forward in the curriculum improves learning outcomes by 0.1 per cent. You can't measure that with 50 students. You need tens of thousands of students to see those differences: 0.1 per cent is not a lot. But if you do 10 of these changes you have 1 per cent improvement; if you do 100, you have 10 per cent improvement. That's big. These are the types of things that you can't do offline.
We're in a position to discover how people learn on a much larger scale, and we're going to release a lot of that information. Let's say moving a single word forward in the curriculum improves learning outcomes by 0.1 per cent. You can't measure that with 50 students. You need tens of thousands of students to see those differences: 0.1 per cent is not a lot. But if you do 10 of these changes you have 1 per cent improvement; if you do 100, you have 10 per cent improvement. That's big. These are the types of things that you can't do offline.
No comments:
Post a Comment