I feel like the initial ID should count for less when subsequent identifications from multiple users argue against it. This happens ALL THE TIME, and it can be an enormous task to correct these simple mistakes in taxa that are poorly curated on here.
“The threshold is >2/3, this one is at 2/3”
So in rarely identified taxa, like grasses, the standard is (in practice, not in theory) 3/4. Unrealistic, but at least I know what’s going on.
Obviously, I personally would like the standard to be 2/3 or more, not more than 2/3. Sigh.
the reason we set this threshold to >2/3 was so that in the very common sequence of IDs:
user a: incorrect id of species 1
user b: incorrect id of species 1
user c: correct id of species 2
the observation would be rolled back to the common ancestor of species 1 and species 2 as opposed to being research grade at species 1
But perhaps you can have an exception for…
user a: incorrect
user b: correct
user c: correct
where user a doesn’t count against the threshold needed to reach research grade
This is also germane to the topic I brought up recently:
…wherein user A (who had opted out of community ID) and user B had misidentified and users C-F provided the correct ID. Since this wasn’t above 2/3, user A’s original misidentification was showing up on the map for the incorrect species and could apparently only be remedied by either recruiting more identifiers or converting some of the misidentifications. Again, for a taxon where few contribute to curating and identifying, this can be burdensome.
I dont feel that there should ever be an assumption built into the calculation that an identification is wrong, or that it is correct. If it needs to be outvoted so be it.
Statistically, though, it seems far more common for the initial ID to be wrong than for multiple subsequent identifiers to be incorrect. Maybe iNat can put a number to that, but I’d guess that its perhaps a 50:1 ratio, maybe more. So if two identifiers are all that’s needed to reach research grade in the absence of misidentifications, I don’t think that a statistically likely misidentification from the observer or initial identifier should count against the efforts of subsequent identifiers. There’s undoubtedly a large backlog of observations that would benefit from this.
And, of course, if the user responsible for the first ID is in fact correct in their determination, they can always recruit additional users to support their contention and swing the vote their way.
Or 2/2, which I encounter a lot.
@loarie, I can see the problem with observations where observations go wrong, wrong, right.
That wrong, right, right, RIGHT pattern is making the City Challenge IDs burdensome to clear. Especially if the original wrong never comes back to put it right.
Maybe count subsequent ID’s as 1.001, and first ID’s as 0.9999 (behind the scenes) so the common wrong, right, right pattern = more than 2/3 right?
Or tag in another identifier and make it wrong, right, right, right. But if expertise is lacking to the point where you can’t find someone to tag, or if no-one has confidence enough to support your ID (adding weight), then perhaps genus is where it should be sitting anyway!
I kinda like that idea. If the first IDer still felt that the subsequent two disagreeing IDs were wrong, they could just override their first ID with the same ID again, with 1.001 weight this time, and then the “regular” ID weighting would be back in play.
With wrong, right, right, right there is 3 chances for the wrong identifier to receive dialogue as to why it is wrong, but with wrong, right, right there is only 2 opportunities
I’ve seen right, wrong, wrong as well (because I was one of the wrongs). Anything that makes a-priori assumptions about right and wrong is fraught in my opinion. Let the community sort it out. If it takes one more vote, so be it.
Moved this from its original thread in Bugs.
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.