Change wording used by the system when downgrading an observation to an higher level taxa

The difference is that it might still be possible for it to become Research Grade at genus or family level as a photo. For a lot of insects and other arthropods it’s not possible to get to species level from photos but genus is easier and still valuable.

How about a more equal “nobody can, and we’re all disappointed together”?

3 Likes

I guess it depends on what you are looking at and why but for that example in particular, Carex is a huge and diverse genus… an unverifiable Carex stricta is far, far more useful than a verified Carex sp., and when you explicitly disagree now, you lose the former. In unverifiable observations, you can still also filter by genus and see it, albeit not research grade.

I know for other taxa that is different, thus the challenge. My issue here is mostly that i don’t want to lose my Carex stricta data point due to some overzealous person who can’t identify it. And yes that has happened before with Carex species and others.

2 Likes

This is my main point… It is not your decision to revert the ID to a higher level, no matter what your expertise level is… The CID algorithm does that. Your role is to make an ID that reflects what YOU think it is. Your expertise can be expressed appropriately through a comment.

4 Likes

From the way I see it, expressing what I think it is sometimes coincides with reverting a previous ID… If I understand your reasoning correctly, i should express what I think it is without having any impact on the ID left by the others, and simply explain my doubts or disagreement on a comment. This pretty much reflects the option number 1. And by doing that you will not affect the CID algorithm in any way, unless the other user will decide to respond to and follow the information given by your comment (this doesn’t happen SO often, in my experience, and definitely doesn’t happen in every case).
So, I am not quite sure that’s the way to go. I see this platform as a valuable tool, but only if we are all able to maintain the observations to a reliable level. And this does not only mean giving what you think it is the correct ID; it also means avoiding that a potentially wrong ID sits there when it doesn’t have any reason to stay there.

But maybe I am just struggling to interpretate your reasoning correctly. Sorry if that’s the case.

5 Likes

[Edit: Feel free to skip to the bold text below. I think it’s a simpler solution than my initial suggestion.]

If someone posts a mammal and IDs it as genus Sorex (long-tailed shrews), and I put an ID of Mammalia, the system will ask “Is the evidence provided enough to confirm this is Sorex?” That’s fine.

Currently, option 1 (green) says “I don’t know but I am sure this is Mammals.” That’s not great. “I am sure” adds a standard that just doesn’t apply and is unnecessary. The system’s question doesn’t include any standard of certainty, and there’s no official standard of certainty for making an identification on iNaturalist that I know of. The standard for identification is up to the individual user, and the question asked by the system is simply, “Is the evidence provided enough to confirm this is Sorex?” Therefore, the green option 1 can just be “I don’t know.” Or if we really want option 1 to restate the chosen ID, it can just be “I don’t know, but this is Mammals.”

Regardless of whether the current green option is retained as-is or revised as above, if this option is selected, the language “johnschneider disagrees this is genus Sorex” should not appear below my ID, and the CID algorithm should not take my ID into account. Those things don’t currently happen, anyway, so that’s fine.

Currently, the only other option that appears (yellow box) is “No, but it is a member of Mammals.” That’s not fine. Obviously I think it’s a member of Mammals, or I wouldn’t have entered Mammals as my ID. What the system should be trying to figure out instead is whether I think it could be a member of Sorex.

This is easily done by replacing existing yellow option 2 with something like the following two options (as responses to the question of “Is the evidence provided enough to confirm this is Sorex?”):

  • In a yellow box: “No. Although this could be a member of Sorex, the evidence provided is not enough to confirm.” (If this option is selected, then something like “johnschneider disagrees the evidence provided is enough to confirm genus Sorex” should appear below my ID.)

    and

  • In a red box: “No. This is not a member of Sorex.” (If this option is selected, “johnschneider disagrees this is genus Sorex” should appear below my ID.)

Replacing the current yellow option with these two options would solve the very real problem identified by the original poster.

I just re-read the original post and thought of a solution that I think is simpler than the one I just described.

Say someone posts a mammal and IDs it as genus Sorex (long-tailed shrews), and I put an ID of Mammalia. If the system then asks whether I disagree with the existing ID and gives me appropriate response options, as suggested below, there doesn’t even need to be a separate question about evidence sufficiency.

The question could just be something like “Do you disagree this is Sorex?” and the response options could be something like:

  • In a red box: “Yes.” (If I choose this option, “johnschneider disagrees that this is genus Sorex” would appear under my ID.)

    and

  • In a yellow box: “No, but the evidence provided is not enough to confirm genus Sorex.” (If I choose this option, “johnschneider disagrees that this is genus Sorex” should not appear under my ID. However, I do think it would be helpful for something like “johnschneider disagrees that the evidence provided is enough to confirm genus Sorex” to appear under my ID.)

I guess there could also still be a green box, saying something like, “No, and the evidence provided may or may not be enough to confirm genus Sorex” – or maybe even just “No.” If I choose this option, there’s shouldn’t be anything under my ID, and the CID algorithm wouldn’t take my ID into account.

Also, if I choose the yellow box, should the CID algorithm take my ID into account? I think that it should, although perhaps(?) it should have less weight than if I choose the red box.

Or I guess there could be two levels of question. Like asking about sufficiency only if I answer “no” to the question about whether I disagree with genus Sorex.

Well, these sound like fairly complicated solutions, after all, and I have to go take care of some other things now. But hopefully this has given the people behind the site some things to think about :)

4 Likes

I get lost when we start talking about what the green button is, and what it could be, and red and yellow and green and blue… dang, my head get’s turned inside out!

My hypothetical cases:

Joe Inatter posts an observation, and calls it Agenus specifica. Harry Identifier comes along, and looks closely… sees it is blue, not green and deduces it can’t be Agenus specifica because those are green. Harry puts an ID of Agenus and ticks the box to say explicitly disagree with Agenus specifica. CID becomes Agenus with >2/3 agreement at that level.

Lets say it was a black and white photo, and Harry can’t tell if it’s blue or green, but knows it will be in genus Agenus, just not sure if Joe is right about the species. Harry IDs as Agenus and doesn’t tick the box to say explicitly disagree. Joe would have seen this in real life, and so would know whether it is blue or green, but hasn’t stated which in the description. It could be Agenus specifica, or it might be some other Agenus, Harry can’t tell from the photo. So Harry also comments: “Hey Joe, Agenus specifica are green, but if this was blue, then it would be best to ID at genus (Agenus in this case)”. CID is at species still as there has not been an explicit disagreement, which is how it should be because Joe saw this in real life, Harry is only seeing the photo. Joe may know more than Harry and can tell by the size of the toenails… who knows! But then Joe reads the comment and replies “oh, didn’t know that, yes it was blue!.. thanks for the tip!” and changes his ID to Agenus. Or he might reply "nah, you can tell by the toenails, but I also did see that it was green :) " in which case Harry sees his reply, and comments back “oh, I didn’t know about the toenail thing, interesting… I’ll have to look into the validity of that, but I can agree with you on the colour” and Harry changes his ID to Agenus specifica.

Or perhaps Harry is of the same mind as the scientific community generally, in that NO Agenus can be determined to species by photo, and that the green/blue thing is a misconception that is rife on the internet so gets mis-ID’d far too often! Same deal… it could be Agenus sopecifica, or it might not. Genus level ID with no explicit disagreement, and comment to Joe that “to ID to species it needs to be put through a blender, and the generally accepted position on these is to ID to genus”.

It falls down in nearly all cases when Joe doesn’t partake in the conversation, as there is no way to back up to the genus without an explicit disagreement. The observation “belongs” to Joe, so I don’t think anyone is morally correct to force the change back up to genus with a lie about their own view of what it is. I know observers that have changed their global settings to opt out of Community ID for this very reason.

We are all “equals” in iNat in terms of our IDs, so how would you feel if someone went through your observations and bumped them back to Plantae simply because that is where they feel they can only be ID’d to from the photograph they are looking at? I can just hear the reply now… “Oh, but some of us are more equal than others”…

It is difficult to walk away from an observation that you feel has the wrong ID, and the more expert you are, the more sure you can be about it being wrong… but it is NOT WRONG! CID is the result of the IDs made by all identifiers involved in that observation. If the CID is showing something that you feel is wrong, then what you need to “fix” is the ID of the other identifier. Two wrongs don’t make a right, so they say, so have a go at changing the viewpoint of the other identifier!

If an observation ends up with two IDs at species by two identifiers that don’t engage in dialog with the community, then it becomes RG, but even then, RG is not that big a deal… it’s just an label. Anyone using the data for serious use should be validating the content anyway, and so the best thing you can do is hit the observation with good comments to highlight concerns you have about species level ID for these.

I won’t go into the idea of tagging in (called brigading by some) as it is ineffective in these cases anyway. I do think iNat devs need to look at ways that can be used to address the absent/unresponsive identifiers with their problematic IDs (and there are other situations where they are problematic too) because if they aren’t present to partake in the discussion and change their positions according to the outcomes, then they aren’t really a part of the community, so why should their IDs count toward Community ID. Perhaps if someone hasn’t been active for the last 12 months, their IDs that are inconsistant with the active participant IDs should be removed from the CID calculation. How this would be implemented would be difficult, perhaps when an account becomes 12 months inactive, it is parsed for dissenting IDs at that stage… dunno… gets messy real fast! I only throw it on the table in case it sparks some other ideas :)

4 Likes

I feel like some of this discussion loses sight of the objective reality that:

A) the organism in an observation belongs to a single taxon at a given level (e.g., genus, species)

B) there either is or isn’t enough evidence in the photograph to correctly label that organism with the taxon it belongs to

As an expert in identification of Taxon A, I see no reason I shouldn’t be able to bump an observation up a taxonomic level (e.g., genus to species) if the answer to B is “there isn’t enough evidence”.

If experts don’t have that ability, then there is no ability for the community to override two users who over-identify an observation to species, even if 1000 observers add a non-disagreeing genus level ID. That would be a significant blow to the quality of the data in iNat.

4 Likes

I guess I see the issue as competing burdens of proof:

  • Identifier-A identifies an observation as Species-X.
  • Identifier-B asks Identifier-A to prove (say why) it is Species-X
  • Identifier-A asks Identifier-B to prove it is not Species-X

Who has the burden of proof? In the iNaturalist world, both do, regardless of expertise. That is how great learning conversations get started around identifying observations.

In this hypothetical scenario, if neither identifier can support their case, then neither should be saying that it is or is not Species-X. That is why Identifier-B needs a third option, instead of disagreeing that it is Species-X, to disagree that species-level (or whatever-level) identification is possible for this observation.

This again comes with its own burden of proof, to convince Identifier-A that they have insufficient evidence to call this Species-X, or Species-Y, or any other species. Then it’s up to Identifier-A to adjust their ID, or justify standing their ground.

All this said, it is well to remember that iNaturalist asks identifiers only to “Suggest an Identification” and “Tell us why…” With photographic evidence, we are usually dealing (at best) with a preponderance of circumstantial evidence, not 100% proof either way. Good enough, I say, for what we are trying to do here.

5 Likes

I have to admit I got pretty confused by all the responses here and I’m still not sure I totally understand what’s at issue.

In the hopes of making it clearer (but at the risk of making it more confusing) I’ve made a silly chart and attached it as jpg image.

@michele_m Can you confirm that this diagram captures your issue? (I don’t want to misconstrue it).

My sense is that there may be no way to avoid the problem that @charlie describes (in the first response) while simultaneously avoiding the problem that @fogartyf describes later on.

7 Likes

To address the specific feature request, was this wording suggested here already? Maybe it was in the Google Group…

Rather than “bouteloua disagrees this is Geranium maculatum”, it could be “bouteloua disagrees with the ID Geranium maculatum”

This covers both bases. I could be disagreeing that it’s that particular species, or I could be saying that the ID is unwarranted given the evidence provided.

12 Likes

Or maybe adding an “or” to the response: “bouteloua disagrees this is Geranium maculatum OR does not believe there is enough evidence to confirm this is is Geranium maculatum”

Pretty long winded but it is pretty specific.

4 Likes

suppose there’s an observation with 3 existing IDs:

  1. species Z
  2. genus Y (parent of Z)
  3. genus X (not a parent of Z but in the same family A as Y)

suppose i want to make an ID family A because i know it’s not Z and it’s not Y, but it might be X or W (W = another genus in family A). when making my ID A, i’d like the system to ask me a question for each of the existing 3 (distinct) IDs:

Even though you made an ID at a higher level, is it still possible that this organism is
genus X?(Yes/No)
genus Y?(Yes/No)
species Z? (Yes/No)

i’d like to select Yes for X and No for Y, and i’d like the system to figure out that since i selected No for Y, my answer for Z must also be No. i’d like each of my explicit disagreements to be noted on the corresponding IDs, and i’d like to be able to go back after the fact and change an explicit disagreement. for example, if i decide that it’s still not Z but it could actually be something in Y, i’d like an option to change my explicit disagreement with Y (instead of having to put a new ID for A out there and redoing the whole disagreement process).

i don’t think it’s important to provide an option to explicitly disagree in cases where someone has, say, a species-level ID and you don’t like it because you think nobody can identify below genus based on existing evidence. i think this is a case where comments work just fine.

This thread has become cumbersome, but right near the top I suggested something similar to you that I feel is more concise: ‘ID to X.x species is incorrect or is not possible based on the available evidence" , I don’t see a need to include the name of the person disagreeing, as that is surely obvious.

5 Likes

3 posts were split to a new topic: Disagreeing ID text is persistent once previous ID is deleted

I like having the two options. The wording is really confusing at first, though.

I use one (yellow, I think) to say I know the plant is NOT what is says, but does belong to the higher taxon.

I sometimes use the other (green?) to say I agree with the higher level. I think this useful because I work with difficult taxa, e.g. grasses and sedges, and sometimes I just want to say yes, the genus is right, good work, but I can’t confirm the species.

6 Likes

I tend not to focus on the wording but rather on the function of the button. If I want to leave the current consensus untouched, I click on the green button. If I want to impinge against the current consensus, I choose the yellow-orange button. @andy71’s chart is a solid summary but I choose not to consider the right-most path as an option for me. I can only decide what I think to be correct, not what others might be able to know or not know. For all I know the observer did put the organism in a blender, or, more seriously, has knowledge external to the post that allows them to make an identification that I might believe cannot be made.

I do not decide what others can know or not know, I only decide what I know. That for me is a difference in the way iNaturalist works: no expert, however that might be defined, has a trumping final authority on what others can or cannot know. Change the wording for clarity, sure, but not the nature of the downgrading functionality.

5 Likes

It is confusing and is causing friction. Reading this thread it would seem that we are divided into two camps:
Those who believe one shouldn’t explicitly disagree with an ID to species level X.x, even if the experts will tell you we cannot currently know for certain from the available info that it is that species (because according to this camp it ‘could be’ the sp X.x so how can one say it isn’t? ), versus
Those who say if we can’t know for certain it is species X.x vs species X.y then we must retreat to the certainty of the genus X.sp.
I feel iNat needs to come to a consensus on best practice and then word the disagreement to reflect this. Currently the wording seems to indicate the iNat regards the best practice is to only explicitly disagree if you know for certain it isn’t species X.x.

3 Likes

To me, the wording can be interpreted both ways, which is where it is causing the confusions. But the OUTCOME of the choices and the iNat policy of community ID definitely support that position.

i do think a lot of this comes down to differences in taxa too. For spiders for instance, there are truly cases where literally no one knows (or at least can describe) how to tell them apart without dissecting them, etc. But for plants, there are multiple keys and also characteristics that are not in keys, such that it’s nearly impossible to say that someone else who knows their stuff can’t distinguish one from another. I have my characteristics I use to tell X from Y, and many times i’ve asked how someone distinguished two plant options and learned a new feature identifying a given species i didn’t know about before! For those reasons, except in the case of duress user type blurred nonsense and way out of range impossibilities (saguaros in Maine, etc) I don’t really think people should be using iNat to explicitly disagree on plants if you don’t know if the ID is correct or not. For sure with users who have left the site and a questionable ID it gets tricky, same with cryptic species and such. So there are rare cases where it seems appropriate. But in most cases, it’s better to ask first. Often a person didn’t even know about the second option (for instance Toxicodendron rydbergii) and willwithdraw the ID to genus or give oyu enough info to explicitly disagree after that.

4 Likes

Your summary seems accurate to me. I think it’s all about the “orange/green”?? on the right sight of @andy71’s chart.
There are a couple times I have used the orange button for that scenario to roll back the majority of observations of a particular species to genus. I imagine that would be concerning for some here. :)

One was Eupeodes americanus and the other was Graphocephala coccinea. Both of them are pretty difficult to separate from similar species in their genera. But in both cases they are the most common of the options, so that a field guide to the most common insects of the area or a quick Google search would only show that species. As a result, there were over 100 Research Grade observations of each going into the computer vision and GBIF etc.

One expert on BugGuide has a way to separate E. americanus but a field guide he reviewed says there’s no way to separate it from other species. Since I made the identifications I’ve learned that G. coccinea can sometimes be separated from similar species depending on the quality of photos, so I’m less confident pushing many of them back to genus. I did receive a bit of pushback which I’m thankful for because I learned from it. But I don’t regret changing the IDs because most of them are still unidentifiable, now people are (hopefully) more cautious identifying these species, I’m still around to withdraw my ID if I’m corrected on individual observations, and they’re (hopefully) not messing up the computer vision as much any more. Many were probably correct but there’s no way to know, and based on the users, quality of photos, and reactions to my identifications I think most of the identifications were based almost solely on the computer vision suggestion.

So anyway with all that said, given the comments on this observation suggesting that only very obvious G. coccinea should be identified to species at this point in time, and the number of Research Grade observations that don’t look super bright, what would be the best way to approach these again?

2 Likes