What are people’s thoughts about certainty level in providing identifications? I see that some identifiers submit a suggestion at a genus (or another higher level) and then provide comments saying what species they think it is. Sometimes I found myself thinking, why don’t you just put that species-level ID if that’s your best assessment, rather than leaving it at the genus level? Now obviously there are gradations of certainty from 100% certain to “this is a guess” and I can understand why you wouldn’t want to create confusion by guessing when there’s uncertainty, but sometimes I think identifiers are just being timid and not willing to state an opinion. IDs on INat are called “suggestions”, which itself implies that it’s not a firm scientific claim as if you’ll be somehow shamed or criticized for being wrong.
So, the question for IDers is, how do you decide whether to put forward a suggestion when you’re a bit uncertain? Do you only submit a species ID when certain or do you sometimes submit an ID and admit uncertainty in the comments, or do you occasionally guess a bit and see what happens? Be honest.
I only do when I am certain, which is why so much of my IDing is either coarse IDs of unknowns or organisms specific to my area (and even then, only the well know ones).
Also, I don’t want to constantly be checking my maverick page, so I tend not to ID to species unless I am certain. Moreover, I am at best a generalist and casual naturalist, so I am fully aware of how much I don’t know; others may feel more confident to go to probable/most likely.
For me it sometimes depends on the observer.
I am more confident with a species level ID if I know the observer won’t blindly agree.
(I don’t want an observation to go incorrectly to GBIF)
I am more confident with a genus level ID being a guess, as its likely there is less danger of it impacting data outside of the platform.
When I ID, I’m always quite sure because I focus on what I really know: mammals and birds. However, it has happened to me that I ID an observation (even though it’s from the area I know) but I’m not 100% so I add a comment saying that I could be wrong with the species.
Some other times, people criticise me because the species I suggested wasn’t the right one (even though I identified a mammal or a bird that I thought that that was the correct species) and I add a comment telling them that anyone can make mistakes.
I place an ID only when I am certain. Right, IDs on INat are called “suggestions”, but… two such identical “suggestions”, and voilà, a “research grade”! :)
Most often when I ID at a higher taxon but mention that I may feel an indication to something more specific, it is because of a separation between Hard and Soft traits. I do ID for typical weasels, Mustelinae, so my examples will be for this taxon.
Hard traits are ones that are solidly upheld in that region’s population enough to be relied on as belonging solely to one taxa and not the other. For example, in the Western United States, Long-tailed Weasels show rich yellow/buffy bellies that do not appear at all on American Stoat or Least Weasel. When I have a trait like that and perhaps other supporting traits (even if they are Softer), I feel very certain and will move to species based on it.
Soft traits are ones that I have picked up entirely by looking at thousands of examples and arriving at a gestalt of little characteristics that are not themselves reliable but can give an impression nonetheless. Back to Long-tailed Weasels, it can be difficult to judge size from photos and even observer accounts without a certain measure. But there is a certain way that the ears are proportional to and placed on the head that can strike as a Long-tailed in the absolute absence of Hard traits. These little things can vary with individuals by sex and age classes, but they do help to share as a Soft impression alongside the Hard traits. I try to indicate whether the traits I reference are Hard or Soft to not confuse others into relying on the less reliable ones. I can only be very certain with reliable traits and leave IDs based on those.
Soft traits are useful to share reasonable doubt on RG observations and let one ask the right questions to determine if that gestalt feeling is on the right track. I have moved community ID to be less specific based on Soft traits in addition to lack of Hard traits for certainty, and ask other identifiers to weigh in. It is also important to consider just how reliable perceived Hard traits are by checking on regional populations. iNat is an amazing tool to compare other observed specimens from the same region to see how reliable any Field Mark is. For example, on the East Coast of the US, Long-tailed Weasels overlap more frequently with the generally accepted Field Mark for American Stoat. The generally accepted Hard trait that Long-tailed Weasels must have a tail length greater than half the head-body length is not actually so reliable for those populations. Asking yourselves to review Field Marks for localised populations on a Hard-Soft scale is very important!
For early identifcation, I think boldness is important. If adding initial IDs on unknowns, moving it into ID workflows in spite of certainty/lack thereof is important I think. It needs to get in front of the right people and bold moves will help. After that, I feel agreements should have more certainty since agreements left without intent to go back and review them could just sway the community ID.
I would rather be boldly less correct but right, than boldly add something to gbif at 80% certainty. Everyone will vary a bit here.
And my own human error, i tend to be less bold if i am tired, and more bold if on a roll looking at only one genus rather than sorting unknowns.
I leave notes also just for myself for future or if someone comes back to ask or for others who id same taxa id they have the same gestalt impression. If i agree with my intitial thought on a second look, or another ider “feels” same, we can be more confident to move it along.
It depends. On how well I know that species. On whether I can see what I need to in the pictures.
Then it depends on other IDs coming in, perhaps with a comment pointing out - why it is this - why it isn’t that.
Sometimes I will push my luck - hook line and sinker - to provoke agreement or disagreement. Currently chewing thru some obs that have been languishing for 10 years, I chance an ID, see what happens.
When I am helping with new or missing species, I comment ‘placeholder’ and withdraw later. Don’t want to be on that leaderboard.
I think people assume that there is a consistent relationship between subjective certainty and the probability of being correct. I think that is probably true for a given person in a particular context, but not true across people, nor often for the same person across contexts where they have greater and lesser expertise.
One person’s “I’m certain” is probably another person’s “I think it’s this species but I’m not sure”.
Interesting to hear people’s logic for this. There are so many factors to consider in making a decision on how confident one is. How well do I know the subject, what are the possible consequences of being wrong?
If it is a poisonous mushroom or plant that is easily misidentified, the stakes could be higher. If it is between a native plant and an introduced look-alike (Taxus baccata English Yew or Taxus brevifolia, Pacific Yew in the PNW for example), maybe a wrong ID could result in someone deciding to root out that tree or preserve its life. And then there’s the scientfic value of Research Grade which, because I’m not using the data as a scientist, I don’t know too much about, but obviously having RG observations that are incorrect could cause some problem later.
I most often only give a species ID if I am certain, but thinking about times when I wasn’t sure and put in a reasonable guess. I might have been too tired and not acting my best, not wanting to work too hard and consult a reference, impatient, etc. We are human and sometimes we cut corners and maybe then we learn by making a mistake.
If there’s uncertainty in my mind, I would try to add a comment that mentions some other species as a possibility if I put a species level ID. It looks like Shore Pine / Pinus contorta because of various specific factors and that’s the native species here, but I guess it could theoretically be Scots Pine / Pinus sylvestris because I don’t see the cone or it’s not close enough to really confirm it. Then it’s down to assessing how likely this is to be planted or maybe an escapee. If it’s in an urban area, it’s often better to leave the genus ID because the chance of it being a planted specimen is there.
I definitely think that there are different approaches to IDing and that different IDers have differing thresholds for certainty. Sometimes being bold can be useful, even if it’s ultimately incorrect. I don’t see a problem with this, though I think I probably tend towards wanting to be sure of my ID (I use a mental assessment of 95% sure to add an ID, but one person’s 95% might be another’s 99%…).
Anyways, my only real suggestion is that, if people know they are among the bolder, more liberal IDers who are willing to make “suggestions” with less certainty, that they regularly revisit their IDs (looking for mavericks, etc.) and withdraw/walk back as needed so that they don’t impede Community IDs becoming more accurate.
This seems to me a rather simple question. The real problem is that we do not know any objective taxa that exist in nature. Any scientific taxon is just a concept that assumes some set of unique characters. Which only that taxon has and reliably distinguishes it from any other. If I clearly see these characters in a photo, I can certainly suggest a taxon. If it is not so, or I am not sure about it (blurry photo, not quite the right angle) – I will present a higher level identification and possibly give additional considerations in the comments. And sometimes clear concepts of species within a genus just don’t exist – "the group needs a modern revision.” Among invertebrates, for example, this is a rather common case. Or there are revisions, but I have not checked the correctness of their concepts personally.
But anyway, reliable identification, for me, is something like filling out a checklist. If all the necessary items can be confidently checked, then confident identification is possible. If not, alas, no. And, of course, the “checklist” (concept) itself can be erroneous. The scientific understanding of taxonomy is constantly improving. And the “certain" identifications of today may no longer be so tomorrow. This is normal.
When I do this it is generally because I am not sure but really interested to hear further input. Maybe I’m at the slightly better than 50/50 point. If I make a species-level suggestion in the comments someone will probably correct me with an explanation if I am wrong. If I am right they will simply confirm and click on by. Either way, I learn a little more.
I don’t ID because I’m an expert. I ID because I am learning.
As someone who IDs flies, sometimes it can be kind of tricky to ID species because there are a lot of traits that need to be taken into consideration and if even one is blurry or missing from a photo, it can make ID next to impossible. However, in many cases there is one species that is incredibly common and ubiquitous while the majority of species are quite rare. So using probability it is likely that common species, but you can’t be sure. So I’ll type “This is probably species X but I can’t be sure” or something to that effect.
Sometimes I also worry about assigning a species ID to blurry photos if I am not 100% sure what it is because some people will just blindly agree with your ID, even if you try to add some nuance and put the word “tentative” next to it. So I often will stick to the genus level or higher with such photos just to be safe.
I suppose the “consequence” to me is that the data is packed in a dataset on GBIF. I feel my responsibility as an IDer is to use my ID “vote” at the level that best encompasses the possibilities given the evidence and tagging in other identifiers to decide what level they think is appropriate. I feel we are tasked with continually curating the dataset so our “specimens” hosted on GBIF are accurately IDed. Back to my weasel examples, there are many iNat observations where the most certain level is subfamily Mustelinae! (here is an example with details on why) But, the consequences are only as dire as the lack of community review (including yourself reviewing past IDs). A taxon with little attention might sit at an incorrect ID for a while. Even taxa with a lot of attention can be found sitting incorrect for a while. All it takes is a bit more attention and ID effort and it can be turned around. iNat Community ID is so great since incorrect IDs can be overturned with reason and attention. Specimens in other collections are not so easy to get access or approval to have their ID rechecked.
There are times where others being bold, even if off the mark, opened up discussion about the fine distinctions between the taxa which become useful records to anyone reviewing that observation. I have learned a lot from ID discussions like this and always share explanations when disagreeing so others might reference the disagreement in the future. I really like how kevintoo put it:
The community ID model is so wonderful because it allows anyone from any level of expertise to participate in ID and discussion. We are always able to return and revise our IDs to better match modern taxa concepts or to incorporate new information/evidence as we learn it. I think conversations on observations that express certainty, uncertainty, boldness, caution, and disagreement have been great tools for creating stronger IDers! I would not be the identifier I am without these discussions. (:
Like many have said, I try to stick with things I am certain of, to avoid scenarios of “the blind leading the blind”. Especially since this can all end-up as training data for the image recognition system, I want to only be as specific as I am confident of an ID. For example, I just saw some kind of thistle, I recognized it from the structure of the blooming flower, I double checked images of thistles in bloom, and the teasel which I’ve seen in the area, and I was quite confident it was some kind of thistle. However, since I couldn’t guarantee what type of thistle or related variety it may have been, I went up to the Family. That way, at least someone who is familiar with that family will be able to see it, review it, and possibly identify it appropriately.
Similarly, per someone’s suggestions for non-experts (https://forum.inaturalist.org/t/useful-inaturalist-tasks-for-non-experts-wiki/35034), I have been going through observations listed as Needs ID / Unknown, and identifying them by kingdom. While it might seem silly to identify something as a plant, remember that you’ve just eliminated billions upon billions of other organisms from the other major kingdoms. It’s actually a pretty huge first step in identifying an unknown specimen, and brings it to the attention of well-studied specialists who may be able to give a more specific identification.
Of course, even in cases where I’m being more specific, (as in the case of the thistle) if someone who is a specialist in that family of flowers has a different ID from a commonly misidentified plant, they can then offer a better ID–so I’m not super worried about being wrong, because there’s still a chance for it to be corrected by the community–but I do my best to be correct.
TL;DR I only identify to the degree of specificity that matches my confidence level.
As others have mentioned in other conversations, there’s always the option to go down and tick the box saying that the identification cannot be improved. I’ve encountered a number of instances where I’ve had to delete my own observations after reviewing them, because the app decided to use the wrong lens or otherwise go out of focus when I snapped the picture.
All that difficult imagery that’s out of focus or distorted by a lens flare could lead to unexpected behavior. While I’m sure the data scientists working on the identification system have likely implemented ways to correct for these sorts of images, I would guess that it’s probably best to only submit images with a minimum degree of quality–and in some way flag images that do not meet that standard.