Identifying from life isn’t like looking at a photo. Often none of the photos of the ants are clear enough to determine species and we are stuck with genus.
I wish there were some more official way to indicate that a genus level identification is the best that can be done with the given information.
Sometimes I will see ants tagged with “Lasius niger” or “Lasius neoniger” and I know that the most correct thing would be to pull it back to just “Lasius” … but that feels like I’m taking something away from the person who observed the creature and it’s really fine if they call it “Lasius niger” even if there is a small chance it might be “Lasius brunneus” or god forbid a member of the brunneus species complex that isn’t even listed here for some reason.
Ultimately this issue is a problem with ants in need of revision, the problem in science is just being mirrored here.
Sometimes pulling an ID back from the species level can feel like it’s unhelpful since there are only two or three possible IDs in the genus, but the genus contains a dozen ants and now they all seem equal. Sometimes there is a handy “species complex” to cover these cases, but not always.
Has anyone proposed allowing a dual IDs? Eg.
“Lasius niger” or “Lasius brunneus” … because right now if I know an ant is one of these two I’m stuck with “Subgenus Lasius” which contains at least two dozen ants.
Or what about a negative ID “It’s NOT Lasius neoniger” …
I wish – but iNat’s ‘be kind’ requires us to add a positive ID – It’s A Plant Which the CID algorithm then counts as hard disagreement with everyone who comes afterwards – and now we need FIVE identifiers.
Genus seems reasonable compared to the whole plant kingdom among my problem children.
Realistically, that’s what a genus ID with hard disagreement is…
Yes, there are times when you don’t know what it is, only that it’s not that species, and that’s more difficult - but iNat’s whole system relies on an observation having a searchable ID, so there’s not really any feasible way to say ‘not XXX’ without proposing some sort of alternative, even if it is just ‘animals’ (though that’s typically unnecessarily unhelpful). Saying ‘all we know is that it’s not this species’ leaves open whole kingdoms full of options! :-)
Also yes, I think there needs to be an understanding that sometimes species ID is impossible, and genus (or even family or above) might be the best you can get without a microscope and/or dissection, but I’m not really sure how you achieve that beyond leaving comments to that effect and hoping people learn and understand.
I do this too. And maybe that’s the best practice.
I think in my original post I describe the issue in a way that is relatable to people who have encountered the same problem but it might not be as clear for those who have not encountered this dilemma. So let me give an example.
A creature is posted with three photos. One or two people identify it as Somegenus ubiquitus and these photos look like that species but lack the details to rule out *Somegenus unexpectedus.
*
If I could see the detail I could rule out Somegenus unexpectedus, but the technically correct action is to put in an ID for just “Somegenus” and leave both options open.
Only Somegenus is a massive junk bin of creatures that all look very similar: Somegenus unexpectedus
Somegenus ubiquitus Somegenus vanitynamus
Somegenus anotheroneus
Somegenus psudoubiquitus Somegenus neoubiquitus … on and on.
Now instead of it being tagged with “Somegenus ubiquitus” with a small chance of that being wrong, it’s just in the junk bin. I know, and can leave a comment explaining that it’s either Somegenus unexpectedus OR Somegenus ubiquitus … but, that’s lost for the purpose of search and survey to some degree. Is this “improving” things?
And of course how strict you are is a judgement call. It was all fun and games to just call every little brown Lasius looking ant “Lasius niger” … but then we have a species like emarginatus being tracked and some of us want to know what’s going on with THAT little ant.
A solution could be to allow for a dual ID “it’s either A or B and here is how you could tell.” Such an ID could count for either one and is more in keeping with how I really ID ants when looking at them around the city. If there are more than two valid choices then there could be too much to explain about how you would determine which was correct, so limit it just to the very common situation where you have it narrowed down to two creatures. (or plants… my apologies for not speaking about the plants fungi and others more directly. Frankly I don’t know how people deal with plants and fungi… ants are enough.)
I think that a dual ID would not bring many benefits for research. I just imagine having a lot of data points saying “here was an ant, it could be this or this species”. When I then want to, e.g., map the distribution of just one species, the dual ID would have to be excluded, so there would be no scientific gain compared to having an observation that is research grade but at genus level.
Are such species described as complexes anywhere in the literature? I’m not a curator, but I believe that in such cases, the complex can be added to iNat’s taxonomy and observers/identifiers can ID to the complex.
Yes if there is a group of sister species, they can be put into a complex. Although many times unrelated species look similar, so in that case you can use an observation field.
So much relatable. we need to add a feature request where the app’s CVM tries to id as the person clicks images. then it guides him/her: “more clarity required”, “take images from XYZ angles” or something. I d not have time to ads that feature request, but if you, go ahead add it but don’t forget to credit me :). In general the app should know that if this is an ant, we will need a clear dorsal, lateral and face view. that is necessary. That would be necessary, but apparently this ocmunity hates my comments and feature request and also my obsession with guides (see further below).
Blurry images? That is not just your problem. It’s our problem, the entire identifier community’s problem but not a problem for the observers. Take one image for each one, a specie did does not matter, only the lifelist and observation count does. “Snap”, “Click” one for each, nothing to do with whether it is blurry, just make a milestone. and we can just hope that the app improves, or observers become more aware. Both are unlikely, but the latter especially impossible at a large scale. Yet the observers, who have often not added a single id to someone else’s observation, do not know the problem. Or at most perhaps a few bird ids. Birds can be identified apparently from even the blurriest images (don’t ask me how because idk anything about birds except the local ones but I have seen obs. so blurry that it the checkbox for whether the id can be improved would have be checked before family, were it an insect). Who to blame? No one. So we ask the app. Many observers just make a heapload of observations that would never cross an identifier’s eyes, or perhaps, there were no identifiers to start. We can fix one problem with guides which I have indefatigably been writing about, but there is apparently no one with the time, or even ability to put in any effort. If you talk about guides, be prepared to face the community’s egroegious hate for this useless feature, and just hope that you do not lose your temper while writing replies. You know that you have dropped a bomb when you write these. You wrote it, and now for these few days you know that you will just be defending your opinion. I am working on several guides, but the supports are often only dried twigs of hope, hope that something will happen, when the entire community is just against them. The other one, we need two filters on the identify page under sort by - “popular”, for observations with lots of comments and disagreements, so that we can get to some mysteries, and the unidentified checkbox, which will help us find old (just make sort by “random”) and unidentified observations which we lost in the heaplot. That feature request I’ve submitted.
There are lots of filters in Identify and many more if you are willing to use URL manipulation.
The filters for “disagreement” and sort by “random”, non-RG should perform the functions you want.
The reason you have not gotten the resonance you hoped for regarding guides is not because people have an unreasonable hatred of them. If you read the responses to your thread with an open mind instead of deciding that anyone who doesn’t agree with you is attacking you, you will see that there are a variety of attitudes towards guides and people who prefer other solutions because the feature as it currently exists does not meet their needs. We do not have control over staff decisions about whether continue to develop or abandon features.
I would also gently suggest that your frustration about guides is off-topic for this thread, which is about the dilemma of how to identify observations where there is no taxon available that precisely expresses the degree of knowledge (and uncertainty) of the ID.
I can’t speak for every observer, but if someone leaves an ID on an insect obs at genus level with a comment like G. speciesone/speciestwo I’m thrilled, hell any time any of my arthropod observations gets any sort of ID I’m grateful, even if it’s bumping it up to a higher level. I know there are a lot more arthropod observers than identifiers (especially for the less showy groups) so just knowing someone knowledgeable has looked at one always makes me happy.
Honestly, I feel a lot more appreciation when someone bumps my arthropod ID from species to genus than when they just confirm my species ID. When that happens it lets me know that there are more options that need to be considered, some of which might not be in the CV. And that is very valuable information that typically only experts can provide. So don’t feel bad about bumping IDs from species to genus!
I think genus level is alright. I know very little about insects and ID them using AI vision, which is fine to a point. But it provides me with species. Later, i might research if i have time, and find out that there are like four of them, with the one suggested maybe more probable, but the method of differentiation is a necropsy of sex organs. In that case, is it not logical to just leave it as genus, if i desire not to kill stuff i observe?
There is this “Based on the evidence, can the Community Taxon be improved?” I use that often.
And i love when someone else comes, ID at genus level and explains, it is probably this, but the photos do not show this or that, so it could be this one as well, or “it is impossible to say, because of the sex organ/pollen microscopy/whatever thing”. It is very helpful, much less than agreeing from fear of hurting someone´s fragile ego.
Atleast they should be in the interface and not just in URLs, that would be an improvement.
They prefer other solutions because iNat has not worked up guides, and many of them may not even be aware of its functionality.
If people without any guidance make ids, you can expect the downsides to be discussed. That is where guides some in. Still, I will keep it off this topic for now as you have said.
Also I wish I could flag CVM’s that produce large amounts for misidentifications for re-evaluation. That would be necessary.
The two specific functions you wanted are in the Identify interface.
This is again off-topic for this thread. Neither lack of guidance when making IDs, nor how to reduce misidentifications, nor iNat’s guides feature are relevant to the original post.
Spider IDer here. I understand where you are coming from. It can be frustrating to bump an observation backnto genus when you know it actually can only be a certain set of species within a much larger variety in this genus. I therefore cherish the species complexes a lot, which allow for delimitation and feel more resolved in the end. However, they can only be used if they have a scientific foundation within the literature.
I have seen two approaches on this very platform to taggle an issue similar to the one described.
Using a project. You might want to bump observations back to genus, but at the same time include them into a project that collects certain species groups, in you example Lasius niger/neoniger. In this way those observations will not get lost in the swamp of other Lasius species and they are perfectly searchable if desired.
Using observation fields. I unfortunately do not know a lot about them, but I am sure someone else might explain how they work if you want to know. You can basically set any variable, e.g. an inofficial species complex if you wish and mark those observations accordingly.. they now become searchable as well.
I have done number 1 in similar cases ( marking undescribed but very characteristic species in an area)
There is also, of course, the ‘cf’ observation field, where you can suggest the species it is most likely to be - with the advantage that it is searchable. I use this quite a bit in the situation you describe. In fact, observation field for particular genera can be quite helpful for grouping observations into smaller phenotypic groups.
This is a common problem with tiny organisms. I am familiar with Pheidole megacephala, and I can say that many (if not most photos) are useless when it comes to trying to pin down the species, but people ID it as such anyways.
Now instead of it being tagged with “Somegenus ubiquitus” with a small chance of that being wrong, it’s just in the junk bin.
If we ID as the more common species for most cases of doubt it will make the rare species look even rarer and also train the computer vision model to propose the common species more often for images where it really isn’t possible to differentiate the two. This also influences the geomodel used for the “expected in the area” suggestions.
I highly suspect that the relative rarity of L. neglectus on iNat compared to how widespread and common it supposedly is according to the literature is in part due to it being misidentified as L. niger quite often.
iNaturalist may look like a community from the outside but it resembles a number of overlapping communities once one looks closer, each having a style and practice of dealing with hard to distinguish taxa. Some knock back everything in doubt to genus (or family), some deal with it through comments or both.
I think the best place for hard to ID mosses is to get the most likely taxon but not confirmed unless all the starts align and every feature shows.
In theory most grasses are identifiable from a complete set of clear photos, having the smallest detail 0.1 mm large. I still get escaped commercial hybrids that cannot be identified as they are not described. I see a lot of “this is A or B depending on..” comments on grasses.
Dual ID has its merits but I don’t see it implemented - it would mean a change to the data structure and rewriting a large chunk of the apps. The same applies for negative IDs.
Which the CID algorithm then counts as hard disagreement with everyone who comes afterwards – and now we need FIVE identifiers.