It’s unfortunate that it’s being interpreted that way, but it’s a protocol we came up with a the best way to deal with these to avoid having them all ‘ingested’ into the system as the wrong species.
High-level IDs are certainly about the best workaround available in the current circumstances. My focus was really to try to understand users’ expectations as a guide for deciding on improved site functionality. Making the site handle people’s observations the way they generally would want and expect would seem to be a good principle.
Also, in terms of ‘no researcher’ finding value in something, there are so many different ways people use iNat data, so I’m not sure that’s entirely true, though I agree it is of limited value if not possible to classify beyond high-level taxonomic units.
There’s always some edge case that disproves any absolute statement, but the value of an “Angiospermae” ID for a multi-species observation is so much lower even than separate IDs of daisy and daffodil that it’s hard to argue in favor of this approach unless it’s definitely the best we can achieve.
In terms of what to do with these observations, the iNat team has said on multiple occasions they don’t want to edit other people’s observations…
I think it’s possible to draw a line between editing an observation (modifying a user’s photo, changing their comment text, removing their chosen ID, etc.) and adjusting the presentation of the data they submitted. We’re talking about preserving all the data the user entered, but simply using the DQA plus an inactivity delay to duplicate some elements. Each “child” observations gets one image from the parent plus all the non-image data. Nothing is lost or edited. I appreciate that this might need to be explicitly stated in the iNat user terms, but I think it remains within the spirit of the relationship between users and iNat.
so a different strategy along with a flag or data quality assessment entry would be to cause the entry to focus only on the first photo - grey out the other photos - then allow that photo to be identified. That way at least some of the data can be used and perhaps the location is more likely to be accurate for the first observation noted?
I see the motivation for this approach, but, to my mind, this would reach the same level of (non-)editing as the auto-split suggestion, but would produce a less useful result.