Duolingo is ruined for me!
I used to LOVE Duolingo. I have a 277 day streak and it would have been more than a year had it not been for one day in Poland without net connectivity. This site has changed my life – at least changed how I spend a good deal of my free time.
But a recent change has completely ruined the joy I get from Duolingo. Let me explain: Until very recently I got a very good mix of questions when doing my Spanish from English and German from English exercises. Some multiple choice, some listening exercises, some translation from the target language and some translations to the target language.
But then suddenly, about 10 days ago, Duolingo just stopped giving me any translations to the target language. From one day to the other, I no longer have to translate sentences into Spanish/German (I get maybe one for each 40-50 questions).
These type of exercises were always the most rewarding to me, especially as I got to the more advanced levels. Translating from the language you’re learning is fine enough, but you’ll never learn to form sentences and use the grammar unless you have to make sentences in your learning language.
I have no idea why this happened – A/B test? Bug? Or Duolingo just decided to almost entirely do away with these exercise altogether? I tried to write to support, but no answer. I put a post in troubleshooting, but only one other user reported having the same problem. I have not found any settings that can change with this.
I am so unbelievably frustrated. Duolingo has been a big part of my life in the past 16 months, but now suddenly I no longer get any joy from doing my Duolingo exercises. I do the reverse course in Spanish, which helps a bit, but still doesn’t remove the irritation over this massive change in the original courses.
I know one can’t really complain over a free service, but I guess my frustration is in a way a testament to how awesome Duolingo was before this - and I would gladly pay money to get Duolingo back the way it was. Now it’s just close to worthless for me.
What I don’t get it why would they make a change like this? Would any language learner really be happy with just learning to read and understand a language, but not being able to actually use it? It just doesn’t make any sense to me. Or is it just some mysterious bug that applies only to a tiny minority of users?
I would be very grateful to anyone who can shed some light on why this change has happened.
UPDATE 25/7/15: It appears it might be possible to get more translations into source language by using the general "Strengthen Skills" button rather than strengthening individual skills: https://www.duolingo.com/comment/9707046$from_email=comment_id=9712732
This is likely an A/B test that has been tested for a considerable amount of time. It is likely that few users even notice it, and fewer care (considering that most are on mobile).
If you know how to access the browser console, try typing this:
Mine says moderate, for example. One possibility is that some users are probably set to "low", or "high", and that affects the number of translation exercises they get. On the other hand this could possibly be a bug.
P.S. For an official explanation of general exercises see this post.
Since the idea is to limit translation challenges, assuming the label identifies its purpose, my guess is that duolingo gives the user less translation exercises. So the alternative is more of the tasks where one describes an image, chooses from a drop down, speaks, transcribes, or completes an MCQ.
Good tip. Mine says: "sg_limit_translation_challenges_experiment: "control".
Not sure, what that means exactly, but must be something along the lines of "that dude is not getting any more translations exercises from us".
Thanks for the link also - I hadn't seen this discussion. It may be that recall exercises turn some users away, but I suspect that would be mostly beginners who just don't want to be challenged all that hard. There should be at the very least an option for advanced users to increase the ratio of recall to recognition. I know they try to reduce the number of options in settings to a minimum, but this is such an essential thing in my view.
That means that you're likely in the control group. Generally speaking that would be a user not experiencing any changes to the standard "exercises", but duolingo may use whatever parameters they choose for that.
Personally, I agree with olimo's post there, and so does the Ceo. These lessons/exercises should be adaptive and be based on an individual's success rate. So people who succeed a lot on recognition tasks get more recall tasks, and people who fail excessively get less.
Sorry, but this is not correct. The lack of L1->L2 translations are not caused by the sg_limit_translation_challenges_experiment test.
Internally Duolingo uses the term translation for L2->L1 translations and the term reverse_translation for L1->L2 translations, so sg_limit_translation_challenges_experiment is actually the opposite: An experiment to limit the number of L2->L1 translations.
This is also consistent with the observations made in this thread: A user with the 'agressive' version will get fewer L2->L1 translations, which (as a side effect!) will leave room for more L1->L2 translations.
With regards to which experiment that is actually causing lack of L1->L2 translations, I am afraid that I have no idea. I see multiple possible candidates in the list of A/B test.
Well, it still matches perfectly well with what the OP is experiencing. If there is a severe limit of L2->L1 translation challenges, then the OP wouldn't be complaining about seeing too many of them. Also considering that there are limited exercises in any given lesson, it does affect the number of L1->L2 (reverse challenges).
But I agree that it could be another A/B test causing this, and quite possibly a bug too.
What? No. I do not really follow your logic here. Maybe you didn't understand what I wrote:
The OP is complaining about the lack of L1->L2 translations.
sg_limit_translation_challenges_experiment limits the number of L2->L1 translations, which has the side-effect that it increases the number of L1->L2 exercises.
This is the opposite of what the OP is experiencing!
The A/B test sets the number of translation challenges (L2-L1), but we can't really rule out that the OP is not part of the test, and we can't be sure about what "control" means.
sg_limit_translation_challenges_experiment limits the number of L2-L1 translations, which has the side-effect that it increases the number of L1-L2 exercises.
Well we can't be sure of that either, for all we know Duolingo is using other types of exercises to make up for the reduced translation challenges. Also paulaha88 experience shows that it is entirely possible that people are kicked/added to the test group based on whatever Duolingo wishes to test. This has happened in the past, though one can't be sure if it was a bug or intentional.
P.S. The A/B test relates to limiting the number of L1-L2 translations, it may decrease them, set them to some arbitrary number, or keep them how they were before the test.
we can't really rule out that the OP is not part of the test, and we can't be sure about what "control" means.
What? We cannot assume that Duolingo uses technical terms correctly? Are you kidding me? Do you really believe this?
Well we can't be sure of that either, for all we know Duolingo is using other types of exercises to make up for the reduced translation challenges.
Yes, they fill the gaps with challenges of all kinds -- including L1->L2 translations. I test my hypotheses before I go public with them, so I know this is true!
Also paulaha88 experience shows that it is entirely possible that people are kicked/added to the test group based on whatever Duolingo wishes to test.
Sure, but only ~1 out of 20 L1->L2 translation challenges is new behavior. So either it is caused by a new experiment or the sg_limit_translation_challenges_experiment have been running for 2+ years.
I've found aggressive, moderate, control. and I assume that aggressive means translating to the target language would be more in the exercises ?
flandybandy explained, this is "An experiment to limit the number of L2-L1 translations."
Assuming that's right, an aggressive setting would likely mean very few translations into your mother tongue. So to make sure there are "enough" exercises in a lesson, Duolingo will likely have more of the other types of exercises, e.g. MCQ, fill in the blank, or Translating into target language.
How they balance that is anyone's guess though.
The linked thread suggests that they want to do whatever generates the most traffic, and they do not want to provide options. This suggests to me that they are headed to pleasing the lowest common denominator, the most casual browsers, and so features for the diligent learners will be neglected or dropped.
The linked thread suggests that they want to do whatever generates the most traffic
But "traffic" is not a meaningful measure of success. We are currently seeing a lot of "traffic" on the forums in the form of bored teenagers, but most of them aren't serious about learning and merely piss off those of us who are.
Success is someone completing the tree in whatever language he or she is learning.
Success is every learner who achieves real proficiency in a new language thanks (in part) to Duolingo.
I agree that wider proficiency should, ideally, be the measure of success, but it seems to me an unrealistic measure since DL can only measure what the learner does within DL's own tree. How can DL measure what I understand from a book I read or a podcast I listen to? In effect, the only viable measure from DL's point of view is whether the learner completes the tree, unless it is whether the learner completes the tree in L3 from his new L2.
If that's the case then Luis von Ahn has really changed his mind about something. In Jan 2014 he said: "At Duolingo, we first think about what’s good for the users in terms of learning a language, and then if we can make money off [a feature], then great, if not, whatever. I think the most important thing is to have a lot of users who are learning a language."
Ability to translate and check translations of immersion documents would be another measure, albeit a very rough and arbitrary one, given that there is no guarantee that translations will ever be checked, let alone checked by people capable of accurately assessing the translation.
It sure would be great, in my mind, if they could allow the more serious learners to purchase optional tools, to get better educational opportunities here. Just as they allow the purchase of timed exercises, we could purchse "production" exercises, where all the prompts are native=foreign. It seems to me like that would not be at all difficult technically, and would cater to the more advanced learners.
To be fair to Duolingo, the very diligent learners who have access to other resources are the people who are likely to succeed no matter what Duolingo does. There have been people teaching themselves languages since long before Duolingo came along, after all.
The real world-changing challenge is how to make language learning accessible enough that people can succeed who would almost definitely not have succeeded if their program weren't around.
"the very diligent learners who have access to other resources are the people who are likely to succeed no matter what Duolingo does"
Indeed. And they are also likely to drop Duolingo from their list of resources if it trivializes itself to the point of providing no benefit (or at least less benefit than other resources).
"The real world-changing challenge is how to make language learning accessible enough that people can succeed who would almost definitely not have succeeded if their program weren't around."
Succeeding at mediocrity is not success. Four billion people who can order a beverage in another language--but nothing more--is not a particularly meritorious outcome.
Amusingly enough, Duolingo is one of the worst resources I know of if all you want to do is learn to order a beverage. Most other things seem to start with "Here's how to order a coffee." You can probably get a long way down the tree in some languages only knowing how to say "I don't have coffee," looking sad, and hoping they get the hint if you don't use other resources ;-)
But let's pretend there are a couple of extremes available here... Duolingo could be designed to perfectly teach exactly one person to be completely fluent. Duolingo could be designed to teach four billion people how to order a beverage in another language.
Both of those are ridiculously extreme, and both of them are kind of bad. The reality clearly lies somewhere in the middle regarding how many people the program will reach, and how far it will get them.
Thus far, from watching other people learn, and learning myself, it seems as if Duolingo serves best as a (relatively) gentle onramp into a language. If they can get you to the point where you can (somewhat painfully) fumble your way through a basic novel or a television show, they've given you enough tools to take the learning and run with it.
They're still expanding trees and providing more resources, and that's great. I just can't blame them if they put their priority on getting as many people as possible to that point where, at least in theory, you have the base knowledge needed to stand up and start carrying the learning load the rest of the way for yourself.
Many people have complained about this (and I suspect it is an A/B test, since I haven't had the same problem). You seem to me, however, by far the most upset of them all. I'm sorry to hear that you feel that Duolingo is ruined for you now, but I think that you can still get a lot out of it. Duolingo isn't that suitable to use as your only resource for learning a language anyway. If you want to practice writing in your target language more, nothing is stopping you! There are lots of language exchange sites, notably Lang-8 where you can write a text in your target language and have it checked by native speakers.
Perhaps I am being overly dramatic. I do use other resources as well, particularly LanguageTransfer, Memrise and sometimes also Lang-8 etc. Duolingo has just always been my core tool and the stuff I had to do every day. I would then do other stuff afterwards on the days where I had time. Now maintaining my streak has become a chore without enjoyment.
If this is an A/B test then I am will be extremely disappointed if the feature become permanent. And I would have a very hard time understanding why Duolingo would want to remove translations to the target language, at least without giving the user some customization options to reintroduce the concept.
I really don't think you can come close to learning a language without having to form sentences in said language, so it seems an extremely shortsighted change. It may drive slightly higher engagement from beginners who don't like to be challenged too much early on, but for serious long term learners it is a disaster, in my view.
I'd like to add that I am not normally part of the chorus who complains about any changes or A/B-test. But this is just crazy to me.
I have had exactly the same experience as OP and am no longer being asked to translate sentences into the language I am learning which has essentially ruined duolingo for me now too as it is less rewarding as I am not typing any of the language out from memory and definitely am not learning anywhere near as much. It's very sad as I was really enjoying using duolingo up until a few weeks ago!
I'm hoping they change this too. Meanwhile, at least with German and Spanish you can finish the trees, then do the reverse trees, which will contain much more translation in the opposite direction.
It's great they're constantly improving the site. But I hope that the team will see that this change is not a positive one.
What to do about it.
I absolutely agree. Language learning without generating your target language is not language learning at all. And I ageee, Duo should fix this very significant gap. But what to do in the meantime?
- I created another thread, The Other 51%: Filling In What Duolingo Is Missing. https://www.duolingo.com/comment/8352230 . It got a bunch of replies, and most of them go beyond the usual "do your reverse tree" and "go to a conversation exchange site."
- Participate in the forums for your study language. e.g. I frequently post to the forum for Spanish speakers learning English. Yesterday someone there said that my Spanish was noticeably improved. The "initial condition" he would've witnessed was only 4-6 weeks previous. I credit the change to the time I spend writing on the Spanish forum.
- Get a language study partner. But with a difference. Instead of using the language exchange sites, find your study partner on Duo's forums. As you hang out there, you will notice people who are serious about their English learning. Look for someone who would make a good match, and make the offer. Most would give their right arm for a native speaker study partner.
- See my thread, Spanish/English Ping-Pong Sentences. https://www.duolingo.com/comment/8359370 . That thread is getting old; I plan to post a new one tomorrow. If Spanish is not your thing, please steal my idea and post it for your study language. Or steal it and improve it. We're all in this together.
I am also upset with the recent changes in lesson format and difficulty level. The biggest change came for me sometime last week when I suddenly started getting very short lessons (usually 7 questions) of low difficulty (one or two word translations, a lot of matching, little or no translation into the target language). So, perhaps there are multiple A/B tests going on at the same time? Based on my experiences, I estimate that the two changes together mean that someone completing the tree once will probably get less than half the amount of practice and exposure to the language as they would have before the changes. The fact that they are testing different levels of difficulty gives me hope that maybe they'll make it customizable (or maybe I'm being overly optimistic). I also wrote them providing feedback about these changes, for what it's worth.
Thanks for pointing me to that. So much for that idea. :( But if they really did go in a more adaptive direction then we wouldn't need customizable levels. However, as it is now it is the opposite - the better you do, the easier the lessons are. I did French for English as a review and so needed the hints very seldom and made few mistakes, and the lessons where I used no hints and got everything right were the shortest and had the least diverse question types.
I've been using timed practice for a better review, and that is still beneficial despite the persistent bug where you get the same question several times in a single practice (even when not getting it wrong to start with).
EagerSnap, you quoted Luis, "At Duolingo, we first think about what’s good for the users in terms of learning a language, ... I think the most important thing is to have a lot of users who are learning a language." Well, his "most important thing" is actually two things: A. a lot of users. B. Learning a language.
Most people active on the forums are passionate about learning languages. But I believe they are not a representative sampling of Duo users. Just Spanish-to-English has over 30 million users! I don't know how many millions of English-to-X users are out there. But on the English forums, probably less than 1,000 users are active. Those are likely to be mostly highly motivated.
I think that the majority of people using Duo are different, if they don't need a new language to make a living. When it gets too hard, they'll quit. So we are met with Luis's two goals, more users, and more learning.
- If the lessons are easier, more people will continue in the program, but they will learn less.
- If the lessons are harder, more people will drop out, but the ones who remain will learn more.
I'm sure Duo wants to find the sweet spot, but from where I stand, it looks like they are trying to reach more users, but with less learning.
for what it's worth, I'm sorry to see this change too. I wouldn't say all is lost now, but I was initially attracted by DuoLingo for being so effective at brushing up on languages I learned long ago, and getting me ready to speak the target language promptly - instead of that coming back after having spent a week somewhere (hard on most schedules, impossible if just attending a single meeting in a different language). Duolingo really served me well for that; I'm now also enjoying learning other languages, but production (written or spoken) of the target language seems to be less of a focus in the lessons, which is too bad.
I am bothered with this a bit as well but since I also do the reverse tree I compensate over there.
For me, I think after recent changes the word pair matching exercises are coming up with too high frequency in particular on my mobile device. So, there is actually value doing a mix of both mobile and website exercises.
My expectation is that for someone like me that has finished his tree and/or over a certain level, should get less of these simpler exercises but more of full sentence translations and at least half of those should be to the target language.
A concurrent change.
Perhaps I can provide some more data to your investigation. See this thread: https://www.duolingo.com/comment/8337049 . I posted that April 28. Background: I participate in only one language type, English speaker learning Spanish. All my practice is in the format Timed Practice. I completed my tree in mid-March. I believe tree completion influences your ratio of target language generation exercises. Summary:
- Over the two prior days I had noted the ratio of language generation questions. It was 2% - 2.5%. I didn't specifically notice this as a "new normal," but maybe I was oblivious. Maybe I'd always had a low ratio of generation questions, or maybe it had changed from a high level to a low level just a few days before. I don't know.
- I posted the thread and got a huge response within the first 24 hours.
- The very next day, April 29, my ratio of language generation questions had jumped to 10%, a 4-fold to 5-fold increase (sampling: 60 questions). I'm certain this was a result of the outcry on the thread. (But of course it could be a fortunate coincidence.
- Twice I posted about the increase in generation questions, but nobody replied. Maybe only my ratios changed. Maybe the thread was dormant and nobody noticed my new data.
Eagersnap, my ratios changed April 29. Might your ratios have changed at the same time?
Thanks, I hadn't seen that thread before. I don't know exactly when my change happened but it might have been on April 27th - sounds like around the same time you noticed unusually low "recall" lessons.
Have you checked your current group in the A/B test by using the browser developer console? Perhaps they switched your group following your post?
Been thinking about this since yesterday. Is there any way we can make our feelings heard? Do the staff take notice of the forums at all?
The more I think about it, the more I feel that retention of users does not need to equal dumbing down. There's two ways to make a learning curve less steep: lower the top of the curve (what they're experimenting with now), or lengthen the curve (which means some of the suggestions others have made: more lessons, sub trees, second trees, automatic increases in difficulty setting as you level up, etc.).
What's the point of working hard to finish the tree if I don't have that "wow, I know so much language now" factor?
I decided to try and see if the issue remains unchanged, so I just did five refresh lessons (Christmas, Objects, Places, Numbers, Past Tense). I got:
- 41 Spanish to English translations
- 10 English to Spanish translations
- 42 questions of other types (multiple choice, listening, single word translation etc.)
It seems to me my ratio of English to Spanish has increased slightly, although since I didn't collect data in a systematic way before it might just be my imagination.
I still think about 11% translations to the target language is way too low, especially for someone who has already completed the tree. I could understand if they started beginners at a very low percentage, but for someone at level 19, the ratio should be at least 30-40% in my opinion.
I am feeling the same, but fortunately later I found out only my iPad app version is heavily affected so I change to the web version when I want to practice badly.
I also wrote a post to propose an opt out and survey/feedback option for testing features that make our life miserable. I hope you check it out and support it.
Oh, this is happening to me too. I am learning German and Turkish and I am not receiving enough exercises for translating to German and Turkish.
This is not good because I no more have the opportunity to practice building sentences in the new languages (50% of the value of the course is gone this way).
So far no change. But at least this thread has shown that many users feel the same way as I do, so hopefully Duolingo will take note and increase this into their considerations when they decided whether to make the changes permanent.
In any case I will keep trying every few days and see if my experience changes.
I just gave it a try, refreshing three lessons. I got 16.67% ENG to ESP translations (10 out of 60). When I did five lessons three days back I only got 10.75% (10 out of 93). So yes, it does seem better now, but whether its a consistent pattern or just random fluctuations I'm not sure.
I noticed the translations seem to come mostly towards the need of the lessons. Perhaps we get more of them if we do well on the other questions there by "proving" to Duolingo that we are capable of handling more difficult ones?
I just did another lesson and I must say I feel a difference. Perhaps they are upping us now. It even surprises me when I see that I have to translate from English to Spanish. What is even more surprising is the amount of words I seem to remember even though I have not done this in a while. Duo knows... Duo knows!
I though too, but then I just did "Countries" refresh without getting a single translation. I wonder if Translations are blocked for certain lessons. I sawa list in the developer tools about a week back saying "No Translations in [lesson name]" once, with perhaps twenty different lessons listed. No idea how I got it up, tried to make an "ñ" and then hit something else by mistake.
I was thinking that I haven't been having as many L1 - L2 translations; however, I do seem to be having more audio "type what you hear" exercises, and I just don't hear them very well. I've tried using the "turtle" option, and that helps some, but is not totally reliable, either. I have my speakers turned up very high. I don't use a microphone - don't have one.
I agree that translating sentences into a foreign language is satisfying. Have you tried doing English from Spanish or English from German? Or you might even try Spanish from German and German from Spanish. I recently started the Italian from French and enjoy it much more than Italian from English since it give me extra practice in writing French.
Yes, I am doing the reverse Spanish course. These days I am doing pretty much only that one. So at least there's that. If the issue doesn't change I might start up the reverse German at some point.
It's just not quite as satisfying. I like leveling up in my chosen language etc. Now I just level up in English. And the discussions are not as relevant to the kind of issues I struggle with, since the Spanish speakers obviously struggle with different issues when learning English. And switching between the courses is a bit of a pain when the source language differs.
But hey, it's absolutely still better than nothing!
EDIT: The listening exercises are also moot in the reverse course.
I have also been experiencing this--I just now noticed this discussion strain, but it started happening for me a few months ago, too. In my experience it's not just that it's not giving me translation exercises into the target language, the lessons are just extremely short (like 7-11 questions), so it's like, if I get them all right, I never quite "reach" the reverse translations; the lessons end before I get to them.
What I've been doing recently to deal with this is just typing "no" as an answer to all of the questions that prompt me to translate into English, and answering all other questions normally. This usually brings me up to ~20 questions in a lesson (many of which marked wrong) and about 8-10 translations from English to target language. Which is acceptable. But it's VERY annoying to have to do this instead of just using the software normally? I've often wondered why on earth they would mess up something that previously worked so well. It's definitely ruined my motivation for Duolingo in general, too. It's really a shame because it's really one of the better resources I've found for Turkish...and it'd be helpful for Esperanto too, if it would work right.
Honestly, my biggest complaint (and I don't want to complain too much for something I get for free) is the lack of "explanations." What are the "rules" for this (grammar) topic, and how are they applied? I really have no idea what "modal" verbs are all about, so it's more or less a matter of "what did it say the last time I missed this question?"
I suspect I must be in an A/B test for using more audio ("type what you hear") exercises. I suppose it would be very helpful for pronunciation - and it is, up to a point - but there are so many words I don't hear clearly, or miss entirely, so it's been taking me a long time to get through a given lesson. I'm pretty much stuck until they start repeating questions that I've already seen the answer to. I'm afraid it winds up being more a "remembering" of a phrase or sentence than actually hearing it. I sometimes find myself telling the computer - "that's not what I heard!" I have difficulty (in hearing) the difference between gender endings and singular/plural endings. Once again, I guess I shouldn't complain about a free application, when most of the problem is probably my own hearing.
If you go to your settings, you can turn off the audio. I don't know what the state of the Italian audio is now, but there have certainly been a lot of complaints about it in the past. I am doing the French tree, and I long long ago turned off the audio (thus I don't get audio exercises) b/c I found robogirl really hard to understand, and sometimes just plain wrong. After I get through with the tree, I'm going to turn the audio back on, just as an experiment. In the meantime, I've found lots of online resources for pronunciation (as I am wanting to be able to speak French, not robo girl French).
Thanks! I thought of that, but I am getting better at figuring it out. Sometimes if I just get a bit of the sound, I can guess, based on the topic I'm studying. On the other hand, I'm afraid I'm beginning to "memorize" the phrases/sentences - unintentionally, to be sure, and that's probably not good. And one reason I started with duolingo is to learn the pronunciation. However, there are some inconsistencies, and that is confusing. Anyway, I only have four more topics to do after I finish the one I'm on. I may try without the sound later on.
One of the things that I found out recently by listening to Lorena R on Youtube is that Spanish runs some of the sounds together to make one long word. So I looked it up and I found this reference:
I have by no means conquered it and probably needs a lot reading out-loud in order to prefect it.