Learning software in need of a theory of learning

I have a new project that I enjoy and I think my students will enjoy, but I have trouble fitting into any known theory of language learning.

I call it the Random English Grammar Generator, and it started off as a way to challenge myself to relearn Java before slowly metamorphosing over time into a semi-professional Javascript/JQuery/XML hobby.  I say semi-professional because I tell my students about it and encourage them to use it but never give points or even extra credit for it owing to the fact that I don’t know of any theory of language learning that could justify spending time on it.  See, it’s fun for me to tinker with as a former CS major, but as a teacher I have trouble explaining its utility, as it’s certainly not meaning-focused, interactive, or communicative.  It’s not even very useful as input because the software-generated examples are decontextualized and sometimes have very odd collocations (which will be improved in the next update, but will never be completely natural).  In short, it is to English students what one of those Rube Goldberg-looking cat gymnasiums (gymnasia?) is to a typical housecat.

A few things make me feel like this is something other than a pure hobby.  I know some kinds of students, mostly my former students in Japan who loved manipulation of abstract systems and perfunctory tokens, who will enjoy playing with it, and this provides me some comfort.  Many ESL departments at universities and community colleges in California also seem to spend money on software packages which are similarly grammar-McNugget-oriented and only slightly less contrived in their examples, and they may show an interest in something like this if I can make it a bit more tailored to the grammar books I know they use (for instance, by putting all the passives in one place and the hypotheticals right after the basic if clauses).  For the moment though, it is a showy jalopy that I spend a lot of time working on but can barely get me to the supermarket.