The Automaton and the Artisan


I make bread for our family about once a week, usually two or three loaves. That’s not much, but since I do it all by hand, I can adjust the ingredients as I like. Whole wheat or white flour? Honey or brown sugar – or just yeast, salt, flour, and water? No problem.

If I wanted to be more efficient, for $15,000 I could buy an SV-209 automatic bread making machine from AliBaba. We could feed hundreds of people, but we’d lose control over how the individual loaves would come out: the price of efficiency is standardization.


I think of this contrast as the artisan/automaton scale. You can see it in materials development for language learners, too.

The digital age brings all sorts of ways to automate the creation of new materials. Search engines, database APIs, and programming frameworks allow for highly efficient, standardized mass production. An extreme version of this approach seems to be at the heart of products like Rosetta Stone and Duolingo, which apply the same frameworks even to different languages.

But there are drawbacks. Algorithms don’t understand culture, so when studying Russian you may see an picture of an American boy eating a peanut butter sandwich. Automatically generated exercises that aren’t reviewed by a human being can be mystifying or even hilarious: just google “shit Duolingo says” for examples. True, Rosetta Stone does offer tutoring sessions with human beings, and Duolingo provides for crowd-sourced discussions; these are tacit admissions that a purely automated approach isn’t enough.

“American Voices” lies at the other extreme. Instead of vacuuming in Youtube videos or TED talks, I listen to dozens of audio pieces each week before selecting one. Glossaries, exercises, and discussion questions are written by experienced language teachers. Websites with related resources are chosen by human beings, not search engines. I hope this brings a richness to the material that can’t be matched by an automated approach.

The teacher in me is proud of this, but I’ll be honest: the entrepreneur in me worries. Preparing a new story for American Voices takes at least two days. No doubt I be able to reduce that time a little as I automate (!) some aspects of formatting content for the web. But choosing stories, writing the introduction, discussion questions, comprehension exercises, and finding links to related stories are tasks can’t be done in Javascript or C. (Well, the gap-fill exercises could be done automatically - but they shouldn’t. That’s how Duolingo ends up with those silly distractors.)

I'm betting that even though I don’t offer as much content, enough users will see the value in our approach to keep the American Voices project viable.

The automaton vs. the artisan – which approach is better? I’m not sure there is a clear answer to this question. There will always be a tension between the business imperatives of efficiency and scale, and the need for material that is thoughtfully prepared, with an awareness of culture and a human touch.

At the end of the day, language teaching is about enabling communication between human beings. Let’s make sure that human beings stay involved in the process of developing learning materials.

Share This