Based on my forum search, there seems to be a (perhaps shaky) consensus to
not to disagree when evidence is insufficient (e.g., to move a species community ID to the general generic ID) unless you can prove that the ID is wrong, and
not to use the “Based on the evidence, can the Community Taxon still be confirmed or improved?” to move a generic ID to research grade unless in very rare cases where one is absolutely sure the ID can’t be improved (i.e., with published evidence that microscopy or genetics are needed).
I want to question both of these. Assume we have an observation with a single ID to species level, and identifiers who see it judge “not enough evidence” and leave it as is. It may stay in “Needs ID” for a very long time, identifiers need to look at it again and again. The worse thing is that the probability that someone erroneously agrees with the ID increases over time, and many observations that are non-diagnostic at species level may end up “research grade” eventually this way.
I understand that using the “can’t be improved” button makes it more unlikely for experts who can do what others can’t to see it. But these experts can still specifically search for those observations, they are not out of the world. I also see that “can’t be improved” might demotivate some observers, though I do think that 1) being honest is always the better way, and 2) if the ID is at genus level it would move to research grade, which can be more motivating than leaving it at “needs ID”.
Now assume we have a genus within a specific region which is not possible to ID at species level in the majority of observations. Research Grade observations at species level will accumulate over time even when evidence is weak, since identifiers are unlikely to disagree to set it back to genus level (point 1 above). This will necessarily generate many possibly inaccurate (or at least non-reproducible) research grade observations. The more common of the identified species is most likely to be selected as ID for new observations, and everyone using Inat to identify species get’s tricked. Sometimes a species is the most common of a genus in Inat even though it is rare in reality.
When I go through RG observations and I encounter one that I am sure cannot be identified at species level, what do I do? I’m thinking about using both 1. and 2. more freely for older observations, while copy&pasting a standard explanation into the comments each time. For newer observations, I would leave it with a broader non-disagreeing ID + copy&past comment first. This could also be a way to rectify those genus records that are off to a degree so that they skew the statistics (example above). Or should we try to rectify such records in the first place?