Because writing is one of the most effective skills in becoming fluent in a language, it is worth investing resources to make it more powerful. I strongly suggest implementing a translation practice tool alongside the writing exchange that operates similarly to the flashcards. Each “card” will provide a sentence in the user’s known language to be translated into the language being learned (or vice versa). An AI will draw from the user’s bank of words being learned or already known to determine the words of the sentence to be translated. There will also be a difficulty setting that determines the sentence complexity, length (perhaps a whole paragraph if on hard mode), and words chosen (more elusive words on hard mode). An AI will then be prompted to make corrections and suggestions with explanations. I know from personal experience and study that implementing such a tool will skyrocket progress toward comfortable speech in language learning.
I have implemented this using Anki. Whenever I come across an interesting sentence, that uses a new verb or a new noun, or a new piece of grammar, I shove it into my Anki deck. I don’t accept that passive listening is enough, and like you I think output is important. This helps me notice and ingest prepositions and other features that I otherwise would struggle with.
Anki is free on some platforms, and at a modest price on iOS which covers the developer’s Apple app store fees. Be aware that there is a fake Anki that stole the name, it’s called Anki Pro or some such name.
Are you actually using any specific platform/website/app for that? Or is it just your idea for using AI in the future? Sounds great!
No, I haven’t found or developed any AI to interact with the user’s wordset. Still, I can’t see why it wouldn’t be possible (granted, it would only be able to draw from a certain portion of the user’s known words after that number passes the thousands because an AI won’t be able to process that many words simultaneously).