As others said, metadata isn’t always partly or entirely present nor always accurate to begin with when an observer uploads, so it’s sometimes different from what observers should (accurately) add as date/time/location (“input data”). For more examples, I sometimes take mobile photos and crop them on a computer before upload, or upload computer digital microscope specimen photo. Shouldn’t using some digital cameras result in lost/inaccurate metadata too?
Only a % of obs. that look to have metadata removed were done so manually. Of those, only a % (if any) were intentionally inaccurate (which would be hard to prove anyway). I assume many obs. w/ manually removed metadata were done for privacy intention.
A general valid question raised is how verifiable/reliable obs. are and if they should be more, which could be something like a feature request. Related, we could also wonder what % obs. have accidentally incorrect or missing input data. If a request were made/approved, it could maybe work in something like the following ways:
Privately submit metadata to staff, curators, or project organizers to review/check to the extent it’s possible to.
Or privately submit metadata through an automated system process that would return a flag-like notification to observers/staff/curators/projects in cases where further verification is needed (e.g. missing metadata, or when input data differs vs. metadata). Like a form saying “Are you sure location, date, and time are accurate?” with Yes and No options. This could also detect and notify of missing input data.
Re: the separate issue of user privacy, I’d prefer iNat not show metadata alongside photos, and instead store it privately or share only with staff, curators, or “trusted” users or projects (or to make a settings-option to not publicly share it). Note metadata typically has no relevance for ID; for anything project/research-related there are already ways to request more info. (e.g. obscured/private coordinates); and identifiers can already ask observers about input data or question it’s accuracy using Data Quality.
Whatever is done or if nothing changes, there will remain at least a small % of obs. with unknown or potentially estimated input data inaccuracy. Nonetheless, it seems the overall RG-side of iNat obs. (ID accuracy included) remain mostly accurate for research or project purposes, after noting limitations at least. If you have reason to think input data is being faked in specific instances you could inform iNat. Theoretically, some largest concerns would be falsely documenting a new record of a species previously thought extinct, or in a location where it wasn’t previously known to occur.