Why not just make ‘auto accept’ an option?
My course a lot of the time tests the users knowledge of adjective and verb word forms for singular and plural; masculine and feminine; and first, second and third person. The different forms are usually very similar words.
My course is a Romanian language course for English speakers.
The singular masculine form for ‘good’ in Romanian is ‘bun’, the plural feminine form for ‘good’ is ‘bune’.
If the the learner is asked for the singular masculine form for ‘good’ in Romanian, and they think it is ‘bune’ when actually it is ‘bun’, they can still get it ‘incorrectly right’ (if you see what I mean). As soon as the learner types the ‘n’ in ‘bune’, the system gives the correct answer for ‘bun’, even though ‘bune’ was going to be written.
I believe the site works on science/psychology/neuroscience (whatever) based algorithms, so the ‘auto accept’ feature, when considering what I wrote above, seems to leave the science and algorithms behind the site null and void under these conditions