I keep running across misidentified observations that are research grade because one person made a bad ID and the observer blindly agreed with them.
Would it be a too dramatic step to require an extra concurring ID to reach research grade when the observer is concurring with someone’s suggestion? It would keep a lot of bad data from slipping into RG status, (but might keep a lot of other otherwise good data out).
Any alternative ideas for solutions to this problem? More education for users to only agree to ID’s that they can confirm themself?
I like the idea of just literally having a ‘like’ button on an ID, because I think a lot of the people who do this are just using it that way rather than actually even thinking about it as an ID.
I use the iPhone app. I didn’t know how to agree without tipping it to RG. I have stopped agreeing on my observations because of it. It seems a big learning curve on the mobile application. Or I jumped in without reading the rules or advice.
I agree, at least 2 ids outside of the observer.
As a personal rule of thumb I consider most RG obs with just 2 IDers (for non-obvious critters at least) as somewhat tentative. Getting a few more agreements always helps the case - which is why I will often go through existing RG obs and add my opinion (for that’s all it is), even if it’s just another an agreement.
One great thing about iNat is that RG status isn’t forever - it can be rescinded if someone with different (hopefully better) information turns up and disagrees. I hope that there are a lot of eyeballs being passed across even RG obs, and that corrections are happening all the time. I know my eyeballs are doing that work, and it sounds like yours are too. I don’t think it’s worth a technical/process ‘fix’ though.
It was discussed, so quite a few proposals were made, like a new, advanced level of RG or deleting agree button for such cases. I’m not a fan of 3 ids cause I have my own observations where I agreed with an id, sometimes years later, not because I wanted to agree, but because I learnt the id later, but I also stumble upon agreers every day and even when I link guidelines that clearly says don’t hit agree, some people live in their own world and can’t get what you’re saying, so something should be done, though it can be achieved with first ider marking the obs as possible for improvement (which can be unsafe for them too). There’s no easy answer.
I would be very disappointed if the agree button was deleted. Perhaps an observer felt 90% sure before the ID was made, and just needed confirmation to feel comfortable IDing to species. Or perhaps the observer did some research after the ID was made, and before hitting agree.
It would just mean you need to write the id on your own, but yes, the last time it was removed it was back very quickly.
Oh, to add, renaming the button is a hot topic for a really long time.
Was there discussion of removing the agree button solely for your own observations, or was this a global removal of the agree button? I feel like just removing the agree button for a user’s own observations could massively reduce this issue, without the need for changing research grade rules or inconveniencing prolific identifiers.
My $.02: A misidentified “Research Grade” observation seems problematic because it is incompatible with a belief that “Research Grade” means you should believe an observation is correctly identified. The “Research Grade” marking is in conflict with what we believe the “Research Grade” marking means. There are two ways to resolve this: try to change the “Research Grade” marking match our belief, or change our belief to match the “Research Grade” marking. I believe the second option is unambiguously the correct one. “Research Grade” does not, never did, and probably never will mean you should simply trust the identification.
I think this would be unnecessarily punitive towards the majority of users, in order to fix the behavior of a minority. Given the overwhelming number of “needs ID” records already, adding the requirement for a third ID would just increase that even more, and bury the records that actually need identification under a pile of things that need one more agreement to confirm.
The users who automatically agree with an identifier’s suggestion are the same ones who will automatically pick a CV suggestion, or automatically confirm their friends’ CV-identified observation… I don’t think it’d really increase the accuracy.
How about making it possible to specifically search for observations identified via the observer agreeing to a suggested ID? That way those who feel like it could periodically check for erroneous records, but it wouldn’t place too big a burden on the overall ID system.
This is really interesting, not more than two or three weeks ago there was someone asking why people keep annoying them by adding superfluous agreeing IDs, and they were arguing that anything added beyond tipping the observation into RG was unnecessary.
FWIF I agree with you, but I would do away with RG altogether.
Yep, this. It also depends somewhat on the purpose for which you’re using the ID. If you’re just trying to find out of interest what the critter you observed crawling by was, fair enough, and probably good enough on balance of probabilities to be reasonably accurate. If however your life depended on a correct ID it (looking for the snake that bit you perhaps) then you’d probably not want to rely on it too much :) Fitness for purpose is the key. Are these fit for “research?” Depends on the kind of research, I think.
For the kind of research I’m interested in (mostly taxonomic), the observations that are suitable for research are the ones with good pictures of the relevant morphological characters. :-) What ID is on them is secondary. Whether or not they’re “Research Grade” is entirely irrelevant.
There are particular contexts in which “Research Grade” might be relevant for research. Suppose I were doing a study on flowering times in a common plant that is very easy to identify. Yeah, I could probably just pull all the “Research Grade” observations and not worry about it too much.
In that kind of flowering time example, though, you’re basically talking about a kind of analysis that isn’t going to be too sensitive to a few misIDs and a taxon that it’s reasonable to expect members of the interested public to be able to identify reliably. In that context, you might even be able to just pull all observations IDed as that taxon, “Needs ID” and “Research Grade”, and have it work out more or less fine. It’d basically be a question of whether whatever increase in misID rate you get is large enough that it outweighs the benefit of just having a lot more data.
Better access to and ability to filter by different aspects of the identifications has been near the top of my list of things I wish iNaturalist would do for a while… there are lots of potentially valuable uses in addition to this one.
