2/27 Generative Research w/Duolingo
Thanks to Molly, we had a chance to visit Duolingo office and met Myra, a lead community specialist and Aya & Vivian, product designers. We spent the first half an hour to learn more about Duolingo and the recent linguistic theory and did a pilot generative research workshop for the second half of the meeting time.
Learn about Duolingo
Myra wears a lot of hats at Duolingo. Her primary job is managing and collaborating with 350 volunteers worldwide to develop the contents. Besides that, she also develops course tools and works on troubleshooting with engineers, linguistic specialists and mathematicians.
Duolingo works based on data analysis and language learning theory. Although there are many theories about efficient learning, Myra pointed out that direct translation from one to another doesn’t work well. Instead, Duolingo teaches explicit grammar rules and conjugation tables in a certain context with pop up messages. It also supports language communities for every single phrase and sentence, so people can learn a language from their mistakes.
Duolingo’s course tree and fluency level are designed based on the memory model. Basically, even if users finished a course it rolls back in a couple of days, encouraging them to review once again. The interval of review is estimated by memory decay theory. For the same manner, the system also measure the proficiency. For example, if a user don’t forget what they learned a week ago and answer the question properly, the system regards that the user successfully retained their language skill.
Generating Ideas
Myra invited Aya and Vivian, product designers and language learners. It was so interesting that all of them have multicultural backgrounds; Myra is half German, Aya is German/Japanese and Vivian is Brazilian/Swedish. We asked them to create a journey map about their language learning experience and explain about their emotions on each stage.
At first I thought they have more advantages than other learners because they could expose more to their second/third language from their family. However, all of them pointed out that their first living experience abroad was the hardest moment. It’s because their language ability was quite limited to get used the full immersive language experience, so they had to practice more with their host families and friends. It was also painful for them to dedicate their leisure time to learn a language, which is why they came up with the idea of learning languages with fun.
Finally, we asked them to brainstorm and design about their “ideal” language learning experience using “magic” technologies. Interestingly, everybody came up with a couple of same ideas, whereas they showed mixed opinions on a certain technology.
Interestingly, all participants pointed out that the earlier the better; if they could go back to their teenager period and learn their second language again, they would start earlier. They came up with some funny ideas; one of them was designing a ‘virtual nanny’ for young children.
When it comes to ‘sign translator’ technology, each participant had different opinions; while Myra thought it might help people understand their surrounding environment easier, Vivian said that it wouldn’t really encourage learners to try understanding a different language as itself.
Summary
It was a great time to meet language experts and learn more about how Duolingo works. Overall, I also felt that we need to improve our generative workshop protocol; for example, we should give participants more freedom in solution brainstorming. We need to either provide more technology options or creative tools so that people can build their own magic devices.
