I am trying to get more involved with iNat in my home state of Colorado. Especially trying to key in on groups that seem to lack consistent reviewers. What I have noticed are several cases of species where there are many (>25) identifications in an area where species are mis-identified. This is then amplified by people seeing all these observations and thinking that species is present. For example, a look at cottontail observations shows people pretty much freely choosing between Mountain, Eastern and Desert in the Denver metro area. In fact Desert (the most likely species on the plains) is the least observed. Mountain which should not occur in the city looks downright common. A second case is people choosing between Southwestern and Western Dwarf Mistletoe seemingly freely (only southwestern is known from the state to my knowledge). Next post misleading common names…
Is there any advice for these situations? It’s a lot of work to go through each observation, and many are verified despite photos clearly unable to move past a generic ID. Is there a way to bulk move observations to the genus level?
welcome to the forum Nick
Short answer, no. You’ll have to go through them one by one
Identifications must be done one at a time manually. The site offers no automated or bulk tools for doing id’s other than on your own records.
Too much risk of bulk edit wars, vandalism etc.
No, it’s not possible, but with the upcoming changes to computer vision suggestions (by default it will only display species that have been seen nearby - deployed on Android first, then web, then iOS), if you help to clean up an area that should prevent a lot of these misidentifications from happening in the future, especially in situations like the mistletoe one (maybe less so with the rabbits).
It’s gotta be done manually, but if you add it to this list here some folks will help out:
It can be frustrating and daunting to clean up large numbers of misidentifications. Even more frustrating is when you do finally make it through all of them and two weeks later it’s just as bad as it was before (ahem…looking at you, Symphyotrichum novae-angliae).
But computer vision changes will hopefully mitigate this. Besides, it’s always been this way. Just need to sit down with a beer and go through observations one by one.
I hear you! Please post away!
No, I went through this a few years ago with plants (Genus Lupinus). It is a, shall we say, shortcoming of this site. The more people mis-identify something, the more people get the suggestion that an organism is present in an area.
What can you do? Make a Journal post about it, then link to that post on all the observations that are incorrect. Or have a text blurb that you copy and paste. After a year or so, you might catch up.
There will always be more people observing than identifying stuff, and you can’t rely on this site for accuracy, but it is getting better in places where several people try to help. I follow every single observation in my county, and do what I can to identify stuff, but I don’t go back in time until I’m interested in a particular species.
So, yeah do the best you can to communicate with your local observers and don’t worry if you can’t correct everything. There is no bulk way to move stuff to Genus, I don’t think.
Are the computer vision’s species’ location dictionaries locked to when it gets rolled out or are they plastic after? That would be a fantastical dream, but it doesn’t hurt to ask!
Do you know what the Wiki list is trying to accomplish? Is it so the algorithm developers can improve the software, or is it to recruit more people to look at these species?
I think it is the latter for the benefit of the former.
If you’re overwhelmed by the number of misidentifications, make a forum or journal post about the problem, tag a bunch of friends and acquaintances, and we’ll give you a hand (if we know enough to help)
Yep, as others have said, you have to deal with this one by one. A good approach to it is to write out a list of reasonings to refute the ID, perhaps even include a reference to a scientific paper or two. Thus you can just copy and paste this into your identifications, and it can give the user’s some information for them to reconsider their ID choice. How the user’s react is another matter entirely (ranges from blind acceptance to extreme critical thinking), but it may induce some discussion (which may lead to some extra help dealing with misIDs).
Maybe even put together a visual key!
The point is to make the data that the model learns from better so that future iterations of the model won’t make the same mistakes.
Much appreciated everyone. I’ll just start chipping away. Hopfully people will start to see my comments as time goes by. The suggestion for the journal is a good one, I’ve not done that yet.
Thanks for tackling this big project. Those “cut and paste” paragraphs will help.
One more thing to add: If the community ID of an observation is at genus level and you are sure that a finer ID isn’t possible, you can click on “Based on the evidence, can the Community Taxon still be confirmed or improved? - No, it’s as good as it can be” at the bottom of the page to make the observation reach Research Grade and remove it from the Needs ID queue.