The Continuing Deterioration of Duolingo

A few years ago I used Duolingo in combination with a number of other resources to learn Esperanto. I also used it to dabble in a number of other languages. My experience at the time lead me to recommend it to people who expressed an interest in learning another language with some caveats. A few months ago I decided to reassign most of the time I spent on social media to more productive activities. One of those activities was returning to language learning. As part of this endeavor I logged back into my Duolingo account. After a couple of years of almost complete absence (I did log in a couple of times, but never to do more than poke around) I discovered that my small list of caveats has grown.

My previous caveats were mostly related to the varying quality of Duolingo’s courses. Most, if not all (I’m not sure about the service’s flagship languages such as German and Spanish), of Duolingo’s course are created, maintained, and updated by volunteers. This results in courses with wildly differing levels of quality. A handful of courses such as the German and Spanish courses are very good. Another handful of courses such as the Swahili course are notoriously bad. But most of the courses lie somewhere in between.

To briefly illustrate the variety of middling quality, I’m going to highlight four courses: Esperanto (a language I know fairly well), Japanese (a language I took in college), Latin (a language I’m decent at reading and writing, but shit at speaking), and Hebrew (a language about which I know almost nothing).

The Esperanto course is quite good. This isn’t too surprising since there are a lot of passionate Esperantists willing to volunteer their time and energy to create educational material (Lernu.net is a great example of this). The Esperanto course includes extensive language notes, audio that is generally good, and enough content (65 skills) to keep learners engaged. But the course hasn’t received a lot of updates since I last used it. In fact the only content update appears to be the inclusion of skills in the main tree that were originally only available by paying lingots (Duolingo’s original in-app currency, which has been replaced by gems… except when it hasn’t). Popular features in the top tier courses, such as stories, are not available in the Esperanto course and I have my doubts they ever will be.

The Japanese course was awful when it was first released. Japanese uses three writing scripts: hiragana, katakana, and kanji. The initial release of the course taught hiragana and katakana, but taught little if any kanji. I also remember the original audio being variable in quality. However, unlike the Esperanto course, the Japanese course has been improved. Now it’s serviceable and there’s apparently a major update about to be released, which hopefully means the course will become decent or even good. But in its current state it still has some issues with kanji. Periodically the shown pronunciations for a kanji character is wrong in the context of a sentence and the pronunciations are written in romanji (showing the pronunciation using the Roman alphabet) instead of furigana (showing the pronunciation using hiragana). The reason this matters is because most elementary level written Japanese material use furigana and higher level material will still use it for lesser known kanji. It’s better the get learners acquainted with how a language is used in the real world. The current course also lacks stories, but it sounds like that’s part of the upcoming update.

I was excited when I heard that a Latin course was going to be released. Latin is one of my favorite languages and I’ve studied it for years. I wasn’t expecting a lot from the Latin course since Duolingo courses tend to be bare bones when they’re first released, but I was expecting more than what was released. The entire course only has 22 skills and only teaches the present indicative tense. There are useful notes and audio for many of the sentences. The pronunciations in the audio are obviously attempting to replicate Classical Latin. For the most part they do an OK job, but not a great job. Unless more skills are added the Latin course is useless for anything other than dipping toes into the Latin waters. With that said, the foundation is good enough that a better course could be built upon it someday.

So far I’ve covered courses for language with which I’m already familiar. Now I’m going to highlight a course from the perspective of a totally new learner. I decided to try the Hebrew course because I wanted to dabble in a Semitic language. The fact that Hebrew is a one of only a few examples of a successfully revived language also makes it a novelty to me. However, I immediately ran into a major roadblock. Hebrew, like Japanese, doesn’t use the Roman alphabet, but the Hebrew course, unlike the Japanese course, doesn’t teach you the alphabet. If you’re completely unfamiliar with Hebrew and want to use the Duolingo course, you need to first find another resource from which to learn the alphabet. Obviously I can’t comment any further on the Hebrew course because I couldn’t get anywhere in it (and as I said I wanted to dabble, I’m not interested enough to seek out other resources), which is what I wanted to highlight.

My first caveat when recommending Duolingo in the past was that some courses were good, some were OK, and some were terrible. If somebody expressed an interest in learning German, Spanish, or even Esperanto, I had no problem recommending Duolingo. If somebody expressed an interest in learning Japanese, I’d warn them away. My other major caveat was that Duolingo couldn’t be used by itself to become fluent in a language. Years ago Duolingo advertised itself as a tool that allowed users to achieve fluency (it would even rate how “fluent” you had become) in another language. The idea that one can achieve fluency in a language solely through translating sentences and typing out what was said in audio recordings is bullshit. Fortunately, Duolingo appears to have backed off from those historical claims and now prefers the much vaguer “learn a language” slogan.

Those two caveats remain, but now I have a number of new caveats when recommending Duolingo.

One of the biggest changes that was starting to roll out when I was first using Duolingo was hearts. Hearts are akin to hit points. Each mistake you make deducts one heart and if you make five mistakes, you’re kicked out the current lesson and blocked from doing anything other than practice. Duolingo claims that the heart system exists to discourage users from making mistakes, but this claim doesn’t hold up for two reasons.

First, what qualifies as a mistake is poorly defined and that definition changes. For example, missing punctuation normally wasn’t considered a mistake. Now it is (at least on some course). Sometimes a typo isn’t counted as a mistake (instead it’s highlighted as a typo, which doesn’t cost a heart), sometimes it is. Second, when you do something that is correct but the volunteers who created the course didn’t anticipate, it gets marked as a mistake and costs a heart. Consider the Latin course for a moment. Compared to English Latin has a very free word order. The standard word order in Latin is subject object verb (which is the same in Japanese, but the standard word order in English is subject verb object). When the Latin course was released on Duolingo a lot of my answers were marked as incorrect because the volunteers apparently assumed that everybody would use subject verb object word order whereas I normally use subject object verb word order for Latin. Likewise, Esperanto has a freer word order than English. Sometimes I’ll provide answers on the Esperanto course in subject object verb word order just to keep things interesting. The Esperanto course has existed long enough where most of those unanticipated answers have been discovered and are now accepted. However, when I first did the Esperanto course, that wasn’t the case. I’ve managed to block myself from progressing in both course by giving correct answers that the course creators didn’t anticipate.

If you run out of hearts, you have a handful of options. First, you can do a practice session, which gives you a single heart. Second, you can wait several hours. You get one heart back after five or six hours. So it takes almost a full day to get all of your hearts back. Third, the Duolingo app periodically provides you the opportunity to regain a heart by watching an ad. Fourth, you can pay gems (but not lingots for reasons I’ll get to in a bit) to get some hearts back. Finally, you can bypass the heart system entirely by signing up for Plus. The hearts feature brings one of the worst aspects of free-to-play games to the educational market: the choice between paying real money or grinding. But Duolingo manages to make this already annoying model worse by punishing you inconsistently and sometimes when you didn’t even make a mistake.

This leads me to one of my new caveats: if you plan to use Duolingo seriously, you should consider either paying for Plus or using the website. What do I mean by using the website? The hearts system only exists in the iOS and Android apps. If you log into the website to use Duolingo, you don’t have to deal with hearts (for now). This brings me to my second new caveat.

Your experience on Duolingo can be significantly different from other users. There are two major reasons for this. First, as I already mentioned, the website experience differs from the experience on the Android and iOS apps. The hearts system isn’t the only difference between the two. Notes that are available on the website can’t be accessed from the phone apps. Without notes you have to resort to a lot of trial and error, but the hearts system punishes you for using trial and error unless you subscribe to Plus. I also made a quip about gems replacing lingots except when they haven’t. If you use the website, you use lingots. If you use the phone apps, you use gems. There isn’t even a one-to-one ratio between lingots and gems. As I type this I have 3310 gems in my iOS app and 954 lingots on the website. When I earn lingots on the website, the number of gems that appear on my iOS app goes up and vice versa, so there is an exchange rate, just not an integer one.

The second reason your experience will vary from other users is A/B testing. Duolingo is infamous for it’s A/B testing. A/B testing is a method where a service provides one experience for one set of users and a different experience for another set of users. Because of Duolingo’s obsession with A/B testing, I have to warn anybody to whom I’m recommending the service that the experience I’m recommending may not be the experience they get. For example, a current A/B test on Duolingo is locking skill tests behind lingots (or gems). If you’re not part of this A/B test, you can test out of skills instead of drudge through multiple lessons. This is useful if, for example, you’re starting a course for a language with which you already have some familiarity. I tested out of the hiragana and katakana skills when I started the Japanese course because I learned those scripts in college (I didn’t test out of other early skills because I wanted a refresher). Since there is almost nothing to buy with lingots, this wouldn’t be a big deal. However, a new user won’t have any lingots so they will have to grind for some before they can skip a skill. If I had been a new user when I started the Japanese course, I would’ve had to do the hiragana and katakana skills, which would have been a waste of my time.

My third new caveat is related Duolingo’s gamification. Gamification is a two-edged sword for educational tools. On the plus side gamification encourages engagement. A user may continue using the app and therefore learning because of the game elements. On the con side gamification often encourages the game aspect of the service over the educational aspect. Duolingo has leagues and leader boards. When you complete a lesson, you get experience points. At the end of the week the top three user in the league win. Mind you the prize is just mostly useless lingots, but that’s enough for a competitive person. This has lead a lot of users to grind experience points in lessons that they can complete with confidence quickly in order to climb the leader board. Since you receive the same amount of experience points for doing a previously completed lesson as you do for a new lesson, there’s no motivation to push yourself in order to win your league. So my third caveat is that if you’re a competitive person, Duolingo may distract you from actually learning.

Rather than improving, Duolingo has gotten worse since I last used it. I used to enthusiastically recommend it for a lot of people. Now I’m hesitant. If somebody is willing to primarily use the website or pay for Plus, it can be a useful service… so long as the language course that interests you is decent and you don’t get trapped in a bad A/B test. What worries me the most is that I see no indication that Duolingo is going to turn itself around. How many headaches will users tolerate for a supplemental tool?

Leave a Reply

Your email address will not be published. Required fields are marked *