i think there are some ways that you could create something that indirectly benefits the underlying machinery of iNaturalist. in other threads, we’ve occasionally brought up the idea of implementing a CAPTCHA-like module that would be based on something like iNat’s Identify page. one direction you could go with this is to bring in Research Grade observations that differ from iNaturalist’s computer vision suggestion. a user of this module would get the observation and would be suggested the RG taxon and the CV-suggested taxon, and he could pick one or the other or his own suggestion. so you get to practice on something that looks like a Identify page, and you can use the module as an actual CAPTCHA and/or accelerate the collection of stats on computer vision discrepancies.
that said, i generally agree with what schoenitz said: