Reading comments here, it seems to me that Duolingo is based on a small set of common misunderstandings, or potentially a set of predication, similar to the following:
1. Universal Grammar gotta be a thing.
2. Perfect parallel texts must always exist.
3. From 1. and 2., "learning a language" must be about projecting an existing UG-to-language matrix into the new one. It's just the matter of how to do that cheap and accurate.
( 4. Also, from 1., any pieces of text must be complete and free-standing, that translation should always be possible and accurate for texts of arbitrary lengths. It's just sometimes complicated. )
But languages don't work that way; sure you can take a C++ program and port it into bare C to JavaScript or Rust, but try between C-like and functional languages and it falls apart. And that's just for artificial language. Language, in reality, is more like a programming language, or like an image codec, heck, like a codec. That's where/how things like these apps as well as machine translations do not work; because UG is not real, because there aren't parallel texts for everything, and there's no such conversion matrix, and you won't even be thinking the same way looking at a same thing in two languages.
Personally I think Duolingo is a great introductory tool and a help, but only to go from nowhere to an overlap region where an illusion of 1. and 2. exists. That region sure exists but definitely a subset of either languages, and that above predication just cannot be pushed outwards to make it a thing. It's just a convenient illusion.
Personally I think Duolingo is a great introductory tool and a help, but only to go from nowhere to an overlap region where an illusion of 1. and 2. exists. That region sure exists but definitely a subset of either languages, and that above predication just cannot be pushed outwards to make it a thing. It's just a convenient illusion.