While completing various typing tests I notice that keys are always arrange by the same order for the same words on app’s virtual keyboard. It defeats learning in some way because instead of remembering letters you actually start to cheat and remember letter positions on the keyboard. By saying “always” I mean I could miss some cases when it’s not happen but generally every similarity in keys arrangement just pops out into my eyes.
I agree with that…
I also have the impression that the keys given in these kinds of tests are always the same, and that may be even worse than them being arranged in the same way (I mean, a normal keyboard is also always arranged in the same way). The simple fact that we are always given the same “fake” letters (which can also be “…” or " "!) is already a bug for me… I had two very similar short words to learn a few weeks ago, with similar meanings, and the “fake” letters where giving a big clue. I knew that for one of the words, I would be given a certain fake letter, and that for the other one I didn’t have the same fake letter.
Here is an example with two very similar words (“утро” and “утром”) translating exactly the same in french (“le matin”). In this specific case, this helps guessing which answer is expected (so I don’t confirm утро when утром is expected since м isn’t one of the fake letters), but still…
First one:
Second one:
I happen to be a native Russian speaker so I totally understand you, @Onesmy. On the other hand, I doubt a computer algorithm can be sofisticated enough to avoid the cases you’re talking about, i.e. fake letters which suggest right answer. I mean, the program must be specifically taught to “see” these cases and act accordingly. Or else it may be some kind of course/lesson metadata which dictates rules for each question-answer pair, kind of: “when displaying this question, don’t use [these letters] as fake ones”.
My exact complain is shown below (just an example):
So, every time I’ve been asked to type the answer to any given question (διαβάζει καλά in this case), I get letters arranged at the same places on the keyboard. All typing tests behave like this.
In my case, the problem is not that I don’t have “м” as an option in one of the cases (it would actually be impossible to guess otherwise without the “auto-advance” feature, if the other fake letters were really random), it’s that in the first case, my “fake” letters seem to be “ь, ф, ч, с” in both typing exercises (I’m still waiting to have a typing review for these words to confirm that this is still the case), while in the second case the fake letters are again the same in both typing exercises (the fact that they are again “ь, ф, ч” is not really bothering my for now).
The fact that they ended up being similar with similar words is also interesting, but maybe this is just a one time event. I’ll wait until I reach the “вечер, вечером” couple. 
As for what you said, I also agree that this could end up more in learning a pattern than an actual word. And this really isn’t specific to one study session (either review or learning). Both the “same pattern” and the “same fake letters” seem to stay (the following screenshots were taken yesterday and today, ю is the only fake letter). Maybe it’s something in the application cache? It “remembers” the last typing exercise and doesn’t create a new one since it has an exercise ready for use? I don’t have a second Android device to install Memrise on at home, so I cannot really try this out, but the two problems are probably related.
So I can confirm this has nothing to do with caching, as I’ve just installed Memrise into a new device (Amazon Fire) and the keys positions are the same for any given word as on my privary device. I guess it’s caused by bad randomization based on words’ letters or their Unicode codes or something like that.
Also, can it be that positions are calculated once for a lesson or a course when it’s created or updated?
Please fix it, Memrise, 'cause I’m sure this is trivial.