Duolanguages

We should be encouraged to practice the difficult part, not the trivially easy parts. But as of March 2022, repetition in Duolingo is unbalanced in this regard.

Yes, I am riffing on the metaphorical use of “unbalanced” to refer to people who are mentally unstable, although I don’t think this literally applies to the good folks at Duolingo, and hopefully won’t be the effect on users either. Although it certainly doesn’t help.

You see, Duolingo has a lovable feature now that lets you “repair” bubbles (themes) that you have already maxed, either by getting them to “legendary” (blue) or just to the maximum level, 5. When some time has passed, the bubbles appear broken with cracks through them, and you are offered double XP for repairing them. This is a pretty good motivation, since XP is what you need to maintain your daily streak. (Although in my case I have only picked 20 XP per day, so it doesn’t really matter. Still, it gives a sense of accomplishment and makes the process of learning more like a game.)

So far, so good. The problem is that the system for picking bubbles to repair pays no attention to whether those lessons were easy or hard. It makes no sense to repeat “Basics 1” in the Finnish course, or at least not more than once a year, since it teaches things you use all the time later on. Likewise, you wouldn’t want to repeat Hiragana 1 in the Japanese course unless you had been away for a year, because you use Hiragana in almost every lesson for the rest of the course.

Duolingo starts very gently, something I deeply appreciate. But after a while, the learning curve starts rising steeply, especially for languages that are very different from English. This is because as you progress, Duolingo adds not only new vocabulary but also new grammar features at the same time. And on top of that, it increases the length and complexity of sentences. And on top of the top, the spoken examples speed up. I assume this will become a problem at a different time for different people, depending on your intelligence and your background in similar languages.

For me in Finnish, the steep uphill started in earnest around the middle of Unit 2, with the lesson called Coffee. This introduces a couple of grammatical features we don’t have in English, like the two different meanings of “to drink”. (“I want to drink something” versus “I want something to drink”. An Englishman will normally say these mean the same thing, but that is not so in Finnish. “Something to drink” is basically translated the same way as “something drinkable”. So one sentence focuses on the nature of the action, but the other sentence focuses on the nature of the object, and the grammar is different.)

Of the next two lessons, Europe was fairly easy because it concentrates on the new grammatical case, inessive. In Finnish, you add -ssa/ssä at the end of the word instead of the preposition “in” that we use in English (and similar prepositions in other Germanic and Romance languages). So that is odd for us, and Duolingo properly devotes the whole lesson to this, teaching only a few new words, and then mainly names of European cities. This is the best way to do things, I think, concentrating each lesson either on new vocabulary or a new grammatical feature, not several things all at once.

My point (finally!) is that I need to repeat those difficult lessons very soon after I have learned them, and then with gradually longer intervals for a while. But the super-easy lessons don’t need to be repeated, or only every half year perhaps. It makes no sense to get the same XP for repeating the words for boy and girl as I get for repeating obscure Uralic grammar. If available at all, the first should count for 5 XP and the latter for 50 XP. That would make sense. And then move those numbers down the language tree as we progress, so that the newest lessons would be repeated more often and give more XP. This would also fit what we know in the 21st century about the neurobiology of learning.

I am sorry to say that some days ago. there were some changes on the Android app, and not small subtle changes like we usually see. I believe it was made more like the iPhone app, and I mostly dislike it.

The first change was the conversion of lingots, the “currency” of Duolingo, to gems. I know this used to be an iPhone thing, but now it has come here too. Along with this, the in-game “prices” of services like streak freezes has gone way up. A weekend amulet used to cost 20 lingots, which was already noticeable but not excessive. Now it is in the hundreds, so I don’t even consider it. There are more gems than there were lingots, but definitely not tenfold more.

I guess the problem was that people accumulate large sums of these lingots over the years, I certainly did, and Duolingo tried to make “sinks” to drain away some of this. But ironically they went so far that I now don’t use these services at all, except “double or nothing”, which over time actually increases my gem reservoir. Did I mention that I have a streak of over 500 now? It is not my first long streak either, although I think it is the longest so far.

The third change was the return of the hearts, and the disappearance of the dumbbells button. This was a bit confusing, but you can still train your existing skills by pressing the heart icon and choosing “+1 Heart practice”.

The hearts system,  I assumed they had abandoned because it was widely hated. (If you search for “Duolingo hearts” you’ll find a veritable jeremiad of upset users). But it is actually not a bad idea, except for its demotivating, customer-quitting effect. For the remaining users, it is a good way to pace your learning so you don’t run too far ahead. Each time you make an error, you lose one of the five hearts. Once you’re out of hearts, you will need to practice earlier lessons to regain them. (As mentioned, just click on the heart icon in the top right and choose “+1 Heart practice”.)

I don’t really run into this as a problem, since I pace myself. I have been an adult for decades and slowly made such habits. So what I miss about this is having the training icon directly on the front page. (I’d also like a “next lesson” button, for pure convenience, but neither of them is strictly necessary.)

The irony here is that the people who most need an external pacing system are the ones most likely to flee from it and never come back, and also tell their friends to avoid Duolingo.

Those are the changes I have noticed. The website still uses the old system with lingots and dumbbell. It also has more in-depth explanations (well, more than the nothing you find in the apps unless you use the forum buttons actively) but more importantly, it is much harder. This is generally good for learning, but also means you have to set aside more time. Popular browsers for Android not only allow you to bookmark Duolingo, but also to place a bookmark directly on the phone’s front page like an app. So that is an alternative for those who just can’t stand the iPhone features.

I was not much of a fan of the “crown levels” to begin with, and I still feel that way personally. But I have no reason to doubt the claim that they have been hugely motivating to a lot of students. After all, I am not a totally normal person, even among the people who learn new languages.

But there is a problem with how crown levels work now, that I believe is an objective flaw. Correcting it would probably make Duolingo better. As far as I know, there has never been an experiment to check whether this would work.

In the earlier versions, daily training would influence the status of the circles you trained. I mean: Each time you used the “dumbbell” training icon that gives 10 XP, it would select exercises from one or two skill circles. Possibly more, but mainly one or two. And when you had finished your exercises for the day, those circles would have improved. They would be completely gold or at least more gold than when you started. If you did enough exercises, you could keep the whole tree gold, or as far as you had come. I did that with Japanese, and mostly with the others while I studied them actively.

But now, when you do the general training exercises, there is no effect on the circles. No matter how many dumbbells you do, the crown numbers don’t change at all.  To change them, you need to do the specific crown exercises for each skill. This makes no sense. If you translate the same sentences, it should have the same effect, because it has the same effect on your learning.

***

One thing that may work right about this, I am not sure yet, is when you max out a skill circle to crown level 5. It looks to me as if you stop getting random exercises from it in your daily training, but again I am not sure. I only started maxing out the early circles once I discovered the feature to let you test out of crown levels. But I haven’t had any of those I maxed out since, so they seem to be at least more rare.

But the thing is, you should be able to max them out (or close to max, at least) by doing sufficiently many ordinary exercises. Maybe you could stop at level 4 in case there are people who specifically enjoy the experience of maxing a skill and don’t want to risk doing it by accident. But the first four levels should be doable. In fact, when crown levels were introduced, they used the stats from my past dumbbell exercises to establish the crown levels of the skills I had learned. So it is possible, they just don’t do it anymore. Or at least I have seen no hint of it. That is the biggest flaw I see right now.

(Others probably see other flaws, like having to spell English correctly when learning Japanese. But English is my third language, so I am not offended if my English is corrected by a computer.)

I think this is a new feature, at least I have not noticed it before. Probably not new as of this day, but fairly new. It is now possible to level up a crown skill by testing out of its current level.

On the web version, when I click on a skill circle, there is a button with a key. Clicking it lets me test out of that level. It takes about as long as a random exercise and gives a truckload of XP, so it is like hitting jackpot. Well, if you manage to do it with less than 4 mistakes. They won’t make it easy for you, they say. But if you’re a bit down the skill tree, then the first bubbles are probably pretty easy, so why not max them out. Especially if that means you won’t get to see them more, or at least not for a long time.

This is a quick way to get a lot of XP (which admittedly means less than before, since levels aren’t really a thing now that we have crown levels). If you have a high daily goal, it is probably more important – I have only 20, which frankly is a bit on the low side for Japanese. It is just that after my daily quote, I already feel kind of wrung out mentally. But I think I have told you about this already, how I feel mentally exhausted even when I get everything right, something I never experienced with French or German.  Anyway, as long as you have some easy skills that are not maxed, you can get a full day’s XP in the time it takes to do one dumbbell. Also you get a free lingot.

In the past, I could not be bothered to go through the loooong row of exercises to max crown levels in a skill I already found easy. But now I can max them out quickly, so this was definitely a nice discovery.

Duolingo has been quite open about the fact that they experiment on their users in order to find what works better and worse. Some users are given one type of lessons or features, others another. After a while the people at Duolingo look at the results and see if one of the approaches works better than the other. All new features have been tested on numerous users first, and some features that have been tested were then quietly discarded. This is an ongoing process.

But I have never seen them test anything this strange before. You see, after completing the daily quota of exercises to maintain a continuous “streak”, we are rewarded with gems or “lingots”. These can be used to purchase various in-game luxuries like outfits for the mascot, or special hidden lessons, or a day off without resetting the “streak”. I haven’t found them particularly useful, so I have a truckload of them by now. But lately I have noticed something weird.

I am a “plus subscriber”, meaning I pay a sum each month for some very limited privileges. Mostly it is simply that I want to support them financially, but the price is ridiculously high compared to entertainment services that have thousands of movies or millions of songs for roughly the same subscription price or less. So yeah, it is pretty much charity. But one thing we get “rewarded” with is more of those mostly worthless lingots. After finishing the day’s quota, I am presented with 3 chests, of which I can open two. (One for my streak and one as a plus subscriber.) The number of lingots in a chest can vary from 1 to 5, but generally so that each day varies from 1 to 3 or from 2 to 5. For quite a while, it seemed to be random how many I got. Not anymore. I always get at least one with the lowest number.

For instance if one day the chests contain 3, 1 and 3 lingots, I get 1 and 3. If it is 2 – 5- 5, I get 2 and 5. It does not matter which chest I click on: Left and right, left and middle or middle and right. Any combination gives the same result.

This was not always so. I did not notice when it started, but it’s been that way for over a full month now at least. Every Singe Day. And I have no idea what the purpose is, except if it is an attempt to make sure I leave the day’s session with a sour aftertaste. There is no economic reason to limit them: I am not sure if you can buy them for real money, but you sure can’t sell them, and they have very limited use and accumulate quickly through normal use. There seems to be no imaginable purpose to code this new limitation except to tell the people who pay the ridiculously overpriced subscription “We just want to remind you that we really, really don’t appreciate your support, and we choose to remind you of this at the best possible time to make you remember it for the rest of the day.”

I assume they are going to sift through their statistic to see how many of us cancel our subscription and how many quit Duolingo altogether, compared to a control group that doesn’t have this bizarre little insult added to their code. Because, science!