One thing I’ve been tempted to do, but haven’t gotten around to, is filtering for commonly cultivated species that definitely do not escape to the wild (say, marigolds here in New England) and marking every one of them as cultivated. The issue with this is that so many cultivated plants do escape to the wild at least occasionally, and some commonly planted species (like Northern Whitecedar, aka Arborvitae) do grow wild in the region.
That doesn’t get to the issue of educating the observers, but it does clean up the Research Grade observations and help to educate the AI.
The larger question for me is how to educate new observers, not only as to the difference between wild and cultivated, but also about how to go about learning to ID an organism themselves. iNaturalist, for all its many, many virtues, can’t substitute for good, old-fashioned wandering around with a field guide or 12 in hand, figuring out what characters you need to look at to ID spruces or warblers or red-fruited pixie cup lichens. (Yes, I’m showing my age here! Paper field guides?!) Those of us who make IDs could do a lot of that kind of education, at least pointing people to the right resources off-iNat - I’ve certainly had the benefit of that myself - but it takes time and energy, and I, for one, don’t always have either of those.