My 2c worth.
Mistakes:
Sure experts make mistakes - but so does everyone! One would hope that an expert would make less mistakes in their speciality than most, and I would strongly feel that that should apply. But even if it does not, it is a non-argument: if everyone makes mistakes, why single out an expert?
Dont throw the baby out with the bathwater: is a few mistakes by an expert reason to support the alternative: dozens of observations without identifications?
Any expert would appreciate a correction, with note. And would probably add to the notes, or provide a note why the correction does not apply.
Agreeing.
I routinely agree to anyone who I consider to be an expert/specialist (whether professional or amateur). I trust my specialists, and value their contributions and competence. In my experience, where they have erred and been "caught out’ they have graciously acknowledged the input. It is only partially about achieving research grade: it is also about giving weight and support. I would probably still do it if iNat had a reputation system and the expert got 1000 votes per ID and I got only 1 vote per ID. And I would still do it, if iNat had no way of my earning a reputation, or if iNat did but my agreeing would not increase my reputation in that case (e.g. I was the last to agree, and thus would not get “support” from other agree-ers later).
What I cannot abide is when I see an expert’s ID languishing because some clearly novices (in the taxon) have supported an alternative - and almost certainly wrong - ID. Under those circumstances I request other observers to help “push” the ID: I do this unashamedly, even when I know that only the expert could possibly know the true ID. I trust my experts to be professional, accurate and honest about their taxa and their views (e.g. lumpers/splitters, pro/anti subspecies, etc).
Possibly in parts of the world where there are few taxa and lots of experts, then users can view experts as a dime a dozen and their input as just another identification. But in areas with far too many species, and where a single expert exists every 100 years - if you are lucky - they are valued above hen’s teeth. Even getting them onto iNaturalist is an achievement! In southern Africa you can sometimes detect when an expert is active on the site by the increase in observations posted on that taxon.
Of course, we now have a new expert promoted by iNat (above all human ones): the AI! Now there is an expert I do not trust (at present),
I fully understand that other people will have different views: I probably would support a very different view to my current one if I lived in the USA, UK or Europe.
Gaming.
Yes people can game a reputation system (used as an argument against implementing one). But people are clearly gaming the current system, so does gaming matter? This whole discussion is really about ways of gaming the existing system to one’s own end. Part of the issue is that we seem to have different ends, or ideas about what the ends should be. I dont think gaming should be an argument against any other system, unless it is palpably going to fail because of potential gaming (in which case I dont believe anyone serious would propose it in the first place).
[[and yes, one can “game” the system without being aware that one is doing it]]
Part of the problem with many of the proposed solutions (e.g. increasing thresholds, holdout periods, popup windows, etc.) is that they will just intensify the gaming, or redirect it, rather than solve it. All these “solutions” come at a cost - making it more difficult for gamers, makes it more difficult and onerous for regular and honest users. This is not a crises we are addressing - just a very good question! Let us not solve problems that dont really exist.
Research Grade
Research grade matters a lot! For one - It determines what data are shared and available to other systems. It determines what pictures are used for AI. As a regular user of Citizen Science data, whatever index of identification surety exists, is as important as the ID itself. However, what iNat calls “research grade” does not have to be two IDs. It does not even have to be fixed (e.g. it could be 5 IDs for birds, 3 for Proteas, and 1 for Monkeybeetles).
But: what we call it, and how we delimit it, will not change the fact that it will be the holy grail. It will be the level that virtually everyone will want their observations to attain. It will be the level that a specialist/expert would want the observation to attain if s/he posts an ID. It will be the level that a researcher will want the data to be when they extract it for conservation planning, or Red List assessment or range expansion, or any other paper they might be doing.
But does it have to be a threshold? Can it not be continuous? Can it not just be the number of unopposed identifications? So instead of casual > needs-ID > Research Grade it can be a scale of finest level IDs: 0 (grey), 1 (red) to (shading to green) infinity (well, in theory anyway). Then the end-users (GBIF, researchers, etc.) can determine their own thresholds, which they may vary between taxa based on rarity, difficulty of ID, number of experts, importance of the data, familiarity to the researcher, etc.
But it will still be gamed! (Go green! Which observations (of x) have the highest number of IDs … push this one (its mine) …)