1. Forum
  2. >
  3. Topic: Duolingo
  4. >
  5. Repetition Algorithm. Making …


Repetition Algorithm. Making mistakes or peek at solutions?

For the sake of the repetition algorithm, is it better or worse to peek at meaning or words in respect to make a mistake?

When should I attempt to answer instead of peeking at meanings? For the sake of simplicity, let's suppose that by peeking at meanings I have a 100% success rate. Should I peek when I have a chance of answering correctly <50%? Or maybe <75%? Or maybe <90%?

For some time I'm going to try this strategy: I won't usually peek and if I'm undecided I will leave out the word from the translation. So I can read it and remember it the next time.

June 12, 2017



I think it's better to admit you're wrong and learn correctly instead of absolutely guessing and potentially learning incorrectly.


Just from the algorithm standpoint, I think it's better to peek. Then the algo knows to reduce your strength for that specific word. If you get the question wrong, I think it reduces the strength of all words that appear in the sentence.

I'm also wary of, as I think elizadeux pointed to, doing something wrong and winding up with it as an ingrained bad habit I then have to work harder to break.


I try without peeking first. If you fail, you just get the same question later, so it's ok.


Reading your post in detail, I think that what I wrote is not an answer, sorry.

I guess the question is, do you want to get 'points' or learn? I think trying and failing is a better way to learn.


I think the point of having a spaced repetition algorithm is trusting it with estimating our knowledge and telling us what exercises we need. My question is: does the algorithm think it's worse to make a mistake or to peek? How much worse is one of the two compared to the other ones?

It's not for the sake of making poinst, it's for the sake of communicating to the algorithm my level of knowledge.


Not sure the algorithm would recognise peeking, would it? As we're only hovering, not clicking on the word, I doubt it registers. Saying that though, it does repeat the audio, so maybe... I sometimes hover to hear the pronunciation, even if I know the meaning. I'll be interested to hear what conclusion you reach from your experiments.

If it does recognise peeking, surely it's just going to tell it that we're not sure about that word/subject and need more practise, just as making a mistake would, so I think it probably doesn't make much difference - that's IF it recognises peeking... If it doesn't, it might think we've mastered something that we haven't, and we'll have a gap in our practise. Hmm, interesting.

I want to be prompted to practise stuff I haven't mastered yet, so I've been trying the answer if I think I know it, sometimes making mistakes, and only peeking if I'm really clueless, so I can learn. Sounds like your strategy is the safest for guaranteed practise prompts though. Good plan.


I think I read somewhere that it does register peeking, and that if you peek, the word will lose proficiency faster.


Ah, that's good. It makes sense that if we're clueless enough to peek, rather than having confidence to try, it would be worse from a mastery perspective. Either way, it's going to prompt us to practise, so all good ;-)


The algorithm recognises absolutely everything. It knows every key stroke you take. It knows exactly how long you take to answer. It knows if you check the hints and probably registers how long it is before you check the hint.

My guess is that you will get a word more often if you make a mistake (and not answering would be considered a mistake) than if you peek.


In some cases, new words are explained clearly in the notes for the skill on the website. So, I make sure to read that first when it's available. New words are also listed clearly in the front of each lesson. I don't always look them up ahead of time, but do sometimes.

Guessing on new words - Sometimes, I can also figure out a new word based on context or similarity to other languages. I'd say I'd have to be at least >75% sure to guess on the website and over >90% sure to guess on the app.

There is a greater price to pay for getting it wrong on the app, so I'm less likely to guess or will guess then double-check the hint. I'm not entirely sure how the algorithm responds to leaving a word blank or guessing incorrectly vs checking the hint. My guess is that getting it incorrect or leaving it out will result in more practice than when you check the hints. There's a lot of evidence that repeatedly getting the same thing wrong is detrimental to learning, so it's seems likely that Duolingo is trying to discourage that behavior. Practice doesn't make perfect. Practice makes permanent.

Recalling words I've already learned - If I've seen a word before but am not able to remember it immediately, I make an effort to think about for a minute to see if it comes to mind or I can figure it out. In other words, I don't check the hints immediately in that case.

Checking the hints - For words that I've never seen before and cannot come up with a good guess, I peek. If I'm less that 50% sure or have no clue, then I see no reason to not peek. Why not start learning the new word as soon as possible? Why wait?

If I'm trying to remember the translation and simply cannot after giving it some thought, then I check the hint to let Duolingo know that I need to practice that word more often.

When I check the hint on the website, I make a note of that and write down the word in my target language with either a drawing or translation to English. For a new lesson on the website, it's not unusual for me to check the hints 2 or 3 times. With the app, I'm even more likely to check the hints because there is a penalty for getting it wrong.

But in the end, everyone should experiment to see what works for them. Having a hard and fast rule: I always peek or I never peek wouldn't work for me. It's more nuanced than that.


Thanks, I wouldn't like to wait a minute or two, I'd rather just answer omitting the unknown word and memorizing it from the solution for the next try. I do exercises pretty fast. I think it's ok because the algorithm features spaces repetition, so I will get that word again soon.


Whatever works for you. As for myself, I'm not literally trying to remember the word for a couple minutes - more a few seconds.

Learn a language in just 5 minutes a day. For free.