So an observer added his first id at species level. Then an identificator comes and finds the id wrong. But instead of providing his id at the genus or family level (though he likely could, as sometimes evidenced by comments like “not GenusA.speciesB, but possibly GenusA.speciesC or speciesD”), he prefers to go far away: Angiospermae, or even Planta… I’d expect someone whooften identifies plnats to at least agree to class, e.g. Magnoliopsida; and some families are often easy to recognize. But no.
And now it takes 2 more identifiers to bring it back to a closer level…
Why, but why, do they do so?
2 Likes
OK I’m a (predominately) plant identifier that does almost exactly what you are asking about. I will give you my 2 reasons: Time and knowledge - I am a specialist identifier - in other words I specialize in one genus (Drosera) and one family Orchidaceae) so if an observation comes up identified as say Caladenia falcata but is clearly not in Orchidaceae ~50% of the time I will have no idea what it is and will hence knock it back to Angiospermae - I could probably do some research and narrow down the id but I don’t have the time for that and I know that there are people with the experience in the rite areas that will id it.
16 Likes
iNat needs a better way for us too disagree.
It is not Species A. End of.
iNat forces us to add a ‘random’ ID then decrees de haut en bas that the disagreeing identifier (and all of us) will be punished by Ancestor Disagreement.
Some identifiers follow their notifications for - it’s A Plant, but many don’t. And those are the ones I whine at when they bubble up in the Pre-Maverick project - which has started a new run ahead of the Great Southern Bioblitz.
7 Likes
It isn’t laziness or deliberately being obstructive.
A specialist may not be able to classify something outside their area of speciality. Even as a more-or-less generalist, sometimes I will know that something is not that genus/species but not be sure exactly what it is instead, so it can happen that I push it back to a higher level and it turns out to be in the same family or tribe as the original ID.
Some IDers also don’t realize that the high-level disagree is a disagreement with all the lower taxa, so they pick a level that they are sure is right without being aware of the consequences.
I find that most IDers are happy to edit their ID if I tag them in such cases. This sometimes requires explaining how ancestor disagreements are calculated on iNat.
9 Likes
There is doubt raised, so the rule is that more identifiers are needed. I’m ok with the requirement.
1 Like
I do that quite frequently with Bees. Frequently people ID things as Bumble Bees that aren’t bees at all, so will toss things up to order or class if I don’t know what it is. Also quite common for someone to upload a picture of a plant with no bee on it, I don’t know plants at all, so I just put Dicots and move on. And yes, I have had plant ID’ers get mad about that, but my wasting time trying to figure out a plant ID that will probably be wrong anyway doesn’t help anyone.
8 Likes
It is not you. You disagreed with ‘bumble bee’. End of.
It is the CID algorithm which is designed as a spanner in the works. And iNat working obstructively, as intended - but why?! And why does iNat force us to add a junk ID, which is then turned against us all?!
4 Likes
But maybe it’s better that an observation waits longer instead of being identified incorrectly? If the identifier from the example doesn’t ID at all because they can’t make a lower-level ID then someone else can agree with the wrong ID (there is a problem with unjustified agreements, isn’t it?) and observation goes to RG straight away. And I don’t think volunteer identifiers should be forced to spend extra time to identify observations to the lower level than they feel comfortable with at the moment.
5 Likes
The original poster is not talking about situations where the organism in question is something completely different than the original ID, but cases where it is in the same family or suborder, but an IDer has disagreed at the level of order or class.
For example, in bee world this would be something like a crabronid wasp being initially misidentified as a bee and an IDer enters a disagreement as “Hymenoptera”, which iNat interprets as disagreeing with all ancestor taxa of Anthophila – so essentially everything but Symphyta. Because Crabronidae and Anthophila have a number of shared ancestors (Apocrita, Aculeata, Apoidea), the disagreement is counted against the correct taxon and requires more IDers to override than if the disagreeing ID had been Apoidea (the lowest shared ancestor).
The problem is that avoiding this requires that the specialist has some idea 1) about what the organism in question is and 2) the various relations between families in the taxonomic structure on iNat (something subject to a certain amount of flux).
I try to ID to the lowest level I am sure about in such cases, but inevitably sometimes I end up creating an ancestor disagreement because I didn’t recognize what type of hymenopteran it was. I withdraw my ID, we move on.
6 Likes
I couldn’t understand your line of reasoning. Firstly, a person disagree with an observation that it is a Bee and enters “Hymenoptera”. I’ve checked up the classification of Crabronidae in wikipedia being Hymenoptera/Apocrita/Apoidea/crabronidae. source wikipedia. A person does not disagree with “Hymenoptera”. He/she will have to enter “Insecta” if they disagree with Hymenoptera, and this ok to me. Despite the many steps up to insecta, it doesn’t takes an incredible big number of people to get it down to the species level again, even if the person couldn’t afford the time to assess and leave it as insecta. or the one who inputs insecta can withdraw later when he sees some new developments.
The case you present is different from the one presented above.
To clarify as I see it, the case that causes friction in the smooth flow of higher IDs:
IDer with some knowledge of a taxon is in a hurry, and bumps the ID up multiple levels past the point where their knowledge could narrow, not knowing that doing so would then require more IDers to overcome those extra taxon jumps.
It might be that some people wouldn’t make those extra taxon jumps if they knew it would take two extra IDs to overcome it. They might take the time to not go all the way to plants, but stop at Dicots, even if they had no idea what came below that.
Or it might be that they made some other judgment about their time and knowledge.
For my part, I’ve been narrowing the ID with a note asking, “Can you agree with dicots? I would help the ID go faster.” I don’t know if it’s made any difference. But I don’t assume anything about the taxon-jumper’s intentions or abilities. And sometimes I learn why they made that decision intentionally.
2 Likes
Click on What’s This at CID on each obs - to see what the algorithm is quietly doing out of sight for each ID. It is both cryptic and deceptive. But the blame appears to rest on the identifier, not on the algorithm which is that One Against Them All.
The taxonomy of the two groups I mentioned in my scenario is as follows:
Hymenoptera - Apocrita - Aculeata - Apoidea - Anthophila
Hymenoptera - Apocrita - Aculeata - Apoidea - Crabronidae
You’ll see that they are closely related, branching off at the level of superfamily (Apoidea).
In the scenario I listed, the original ID was Anthophila (bees). A bee specialist enters an ID (disagree, yellow button) pushing it back to Hymenoptera because they know it is not a bee but they don’t know what type of hymenopteran it is.
iNat considers this a disagreement not just with Anthophila, but with all the levels between Anthophila and Hymenoptera (i.e. Apoidea, Aculeata, Apocrita).
Another IDer comes along and IDs as Crabronidae. Maybe at this point the observer withdraws their original Anthophila ID. So there are two IDs, one for Hymenoptera and one for Crabronidae. However, even though these two IDs do not conflict with each other, the community ID remains at Hymenoptera because the original Hymenoptera ID was entered as a disagreement that included the shared parent taxa of both Anthophila and Crabronidae. This disagreement was not intentional – the bee specialist just wanted to get the observation out of bees so that it would be looked at by users who specialize in other hymenopteran families.
So unless the bee specialist withdraws their ID, more IDers are required to overcome the disagreement and get the community taxon to show up as Crabronidae. And then additional IDers who can take it from family to genus or species.
This does matter because there aren’t a lot of hymenopteran experts. There may be 2 or 3 people working on Crabronidae in a particular region, but because of the way iNat calculates disagreements, more IDers are required to overcome the ancestor disagreement.
Our hypothetical bee expert didn’t do anything wrong – they did exactly what they should have, namely, correcting a wrong ID at a level that reflected their knowledge. But the way iNat interprets these disagreements has repercussions that not all IDers are aware of.
8 Likes
I see. Thanks for the explanation.It is the nature that some insects will rarely gets looked into.
Maybe I’ll add a bit more…The new scenario now has two identities registered, Hymenoptera and Crabronidae. The Hymenoptera entry is a rejection of Anthophila, but may include Ampulicidae, Sphedcidae, Crabronidae. It carries a bit more. In a more simple setting, when an user enters the identity to a new observation, it just shows up as intended. It is need ID, requires a single agree to take it to research grade.
1 Like
What would you envision “It is not Species A” doing? Leaving the observation in Species A for now, where it would be misleading (e.g. create a dot on maps)? Canceling out the Species A identification (which would have to shift it to Unknown, it seems to me)? iNaturalist’s current process does cause problems, but if there’s an alternative that doesn’t I can’t think of it right now.
That said, we should of course use the lowest taxonomic level we can when disagreeing. (I’ve seen Crocosmia, a non-orchid monocot, as well as the dicots Lathyrus latifolius [a sweet pea] and Polygala spp. misidentified as orchids so I can understand giving “Angiosperms” as the ID of an observation that isn’t the orchid species it’s ID’d as, odd though that seems at first.) And keeping up with our notifications would really help, much time though that takes if we’re active IDers.
5 Likes
When entered as a disagreeing ID for Anthophila, Hymenoptera is effectively treated by iNat as a disagreement with Ampulicidae, Sphecidae and Crabronidae as well as Anthophila because all of these taxa are within Apoidea.
2 Likes
I think there’s a consideration here that the system works in a perfect circumstance, which is everyone involved jumping back in to either retract their general ID or agree with the later determined specific ID, after this process has carried out. This is great when that happens, but the issue is when people who were involved earlier (and therefore have lingering “general taxa” IDs) are either inactive, don’t return to the observation, or otherwise don’t engage again.
All things considered I think this is one of the best ways to actually facilitate iNaturalist’s identification system, but it is just unfortunate that many times it becomes an impediment due to the circumstances above.
6 Likes
I’ll think in another way. Hymenoptera consists of a few branches/families other than Apoidea, the software can’t name it Crabronidae with just 2 identifiers, each holding a different id. (I mean potentially different ID) I do not know whether the software rejects Apoidea entirely. The other branches of Apocrita are probably in the consideration.
I’ve wondered whether there might be a way to calculate disagreements as a vote only against the current ID and any child taxa, while still displaying the new (high-level) ID as the community ID until some other, more specific ID is entered.
In other words, if the current ID is, say, the plant genus Digitalis and I disagree with it, entering an ID of Magnoliopsida because I’m not sure about the family, it would count as a disagreement with Digitalis and all the species within it, but not a disagreement with the family Plantaginaceae or the order Lamiales (as would currently be the case). The ID would be displayed as Magnoliopsida at this point.
If another IDer comes along and IDs the plant as Veronica (same family) it would be calculated as:
3 votes, all of which are in agreement with Magnoliopsida (2 of these being at a lower taxonomic level; these two lower level IDs would be used to calculate the lowest common taxon)
1 vote for Digitalis / 1 vote against Digitalis
1 vote for Veronica
with the community ID displaying as the lowest common taxon of the conflicting IDs (i.e. Plantaginaceae) until there is a 2/3 + 1 majority for one or the other genus.
…however, I don’t know if something like this would be feasible in practice. It seems like it would make the algorithm quite complicated and potentially even more difficult to comprehend than the current system, given the many different levels at which IDs may be made and conflict with one another.
5 Likes