I’ve been IDing a lot of non-RG observations in a specific taxon lately, and that often entails going through rather grainly images. As a result, the safest ID in many of these is RG at subfamily or genus because there is simply no way to ID them beyond that, the diagnostics just can’t be made out to do so. I’d link examples but I was told to remove the ones I had put earlier, and I think most IDers know what I’m talking about. My concern is that I have noticed a lot of these getting bumped to RG lately, often due to one person coming in and marking as species (almost always with no attached reasoning, and questionable consistency in requests to elaborate) which results in the OP and possible others agreeing with them and sometimes bumping it to RG species.
What should I do about this? It is really frustrating to see such images getting the same level of research-readiness as others that do have clear photos and show all the important characters.
Hey Orrin, I’ve temporarily unlisted this, can you please edit your post to remove any specific references to users, links, etc., and instead format this as a general discussion. I will then relist it, thanks
will do, thanks
@thebeachcomber done!
@thebeachcomber please feel free to delete this comment if it breaks the rule on mentioning specific users, but I do think it may be of benefit to note that the initial species IDs on a lot of these observations are primarily from one user (not to say that it is soley them, but they are certainly a plurality if not majority). I have reached out to said user and after a good deal of healthy debate in DMs, we have agreed to disagree and just carry on IDing as we have been. Ideally I’d like a better long term solution but I don’t know if one exists.
It can be an issue that observers just automatically agree on ID’s with no apparent knowledge of the taxa. Especially annoying when they agree to every new level added. But, that’s why identifying in RG is important.
I have noticed this situation only the other day. Someone uploaded a really far away image zoomed in of a blur sitting on a fence. It wasn’t clear what it could be other than a bird, but they had identified it to species level. I then commented saying it’s not identifiable and asking if they have other photos, but they didn’t. Then other users confirmed the ID and it got research grade despite being impossible to make out. I did think about knocking the ID back to birds, but I decided against it because I didn’t want to rock the boat.
I think what happens is a user might upload a super blurry or tiny image which is undeniable. Then a well intentioned user might notice the observation is super blurry, but still want to give a species level ID. The result is they give a blurry picture a likely incorrect species level ID. You know for certain that the observation was not identifiable to species level, yet it gets research grade anyway. I also think there are users who just agree blindly with ID’s without actually knowing for certain. I think those people are well intentioned, but their accuracy is poor because they go for quantity of ID’s over quality.
I personally wish there was a data quality requirement that could be ticked if the observation has an incredibly poor photo. Often you can see that there is an organism there, but you can’t take the ID past family a lot of the time. Sometimes you can’t even make out what it is at all! In those kinds of circumstances these kinds of observations should have no chance of reaching research grade status as they are basically no better than having no evidence provided. I don’t know how images are selected to be added to the AI’s training model, but I imagine pictures like this have the ability to pollute the AI’s accuracy if the images get added the training model.
Could the Data Quality Assessment, “Evidence of Organism” be perhaps marked to “No”? Unsure if it’s appropriate, but it seems that a blurry image isn’t enough to say that there’s evidence of a particular organism.
In my experience there is two types of lichen experts. Those who only make a species level ID if it’s 100% certain to be that species, and those who ID to the most likely species, even if there’s like a 1% chance it could be something else. The first group will just never ID to species level with a lot of taxons (except if they have things like spore microscopy or chemical test results) - but the second will usually pick a species for my pictures that is very likely. I never agree with any IDs myself, so they do not go to RG, but at least have a species name attached, and I like that for taxon counts and checklists (I know I did see either that common species, or one very rare similar looking one, but for the purpose of species count it makes perfect sense to have it at species - usually if there’s no species-complex made it would otherwise not count as separate taxon at all, unless the genus has only those two species).
If someone goes through my observations and just clicks the Agree button on everything that does indeed push them to RG and I find that annoying as well. The ones I catch fast enough I can block which at least prevents them from agreeing to more IDs, but only works for my own observations of course.
Best modus operandi perhaps a comment:
Not enough detail for further finer identification.
I have done this and it often works.
Don’t be afraid to rock the boat. I’ve done it before if it’s not identifiable you can’t ID it to species.
Someone posted a picture of a raceme of flowers with slits cut into the flower. She says she saw a carpenter bee do it and ID’d the photo as Carpenter Bee.
I ID’d it as Bees because other bees exhibit this behaviour and in the absence of a bee in the picture you can’t really ID it as such even if the person says they saw it.
She was very reasonable and accepted the explanation and we kept it as bee.
You never know what could be a teaching moment.
In this general situation, I think it’s valid to add a disagreeing ID to the higher level taxon that the IDer feels is supported by the evidence with a comment explaining the reasoning. Other IDers can always chime in with their reasoning to explain the more specific ID if they wish.
It’s entirely possible that the observer may have gotten a good look at the organism but a bad pick, so their specific ID may be justified. But IDers can only evaluate the evidence in the observation. Sometimes the observer can add some detail in the description or a comment to support a more specific ID than is makeable from the media, but IDers will need to evaluate that for themselves.
IDers shouldn’t downvote
in this situation as it isn’t what that DQA is for. IDers would use that in a situation where there was no evidence pertaining to the organism at all (like a photo of a rock without any organism or whatever).
If your tentative ID (because blurry and no details) is a comment - then at least it takes 2 to agree. And your comment stands.
If they are only 2, you can add your disagreement back to a broader ID as the third one.
Frankly, for bad photos - I Mark as Reviewed and chose another, which is worth the time and effort.
This can be irritating. However, from a data perspective, to the extent that such observations are just another record of a common species with lots of other records nearby, they will have minimal effect on any analysis done with Research Grade data.
I would suggest to focus corrections on records which don’t meet those criteria: where there aren’t other validated records of the species from nearby in time or space, such as the only record from a county or state of a particular species, or a summer visitor recorded in the winter. In those cases, the burden of proof is higher and there is an additional justification to knock it back to a higher-level ID. But if it’s just another House Sparrow in a city with hundreds of observations of House Sparrows, I would usually just review and move on.
I have seen Identifiers use the “Based on the evidence, can the Community Taxon still be confirmed or improved?” option to mark “No, it’s as good as can be.” to alert me, the Observer, that the Identification on a particular Observation cannot be improved further, and I appreciate it, because it tells me OK, this one need not be pursued further. I wonder if this is appropriate here?
I think it’s important to remember that sometimes even terrible grainy photos are still identifiable, especially when we’re talking about birds. I’ve seen lots of bird observations that someone stuck in “id is as good as it can be” when it was totally identifiable.
Does this carrying on include you regularly making disagreeing IDs that push the RG obs back up to genus/family? It sounds like that would continually be called for, but would continue to be frustrating for both of you… Ideally after enough iterations you could mutually figure out what a good baseline standard of evidence is and not have to constantly be checking each others’ work.
But this is a common issue with many popular-but-difficult-to-ID species, where someone has to constantly keep up with new observations and push all the RG ones back up to the level that the organisms are actually identifiable to.
It would help if we could always choose to put a hard disagreement (one that denies a lower level ID) on any observation. Sometimes I really think the ID was wrong but the combination of previous ID’s prevents the hard disagreement. I can add a higher level agreement and explain why I think it’s wrong, but sometimes a more effective tool would be welcome.
Other times, I’m fine with how the system works, especially when the observation was originally ID’d to subspecies, I don’t know if it’s that subspecies or not, and my adding the species level agreement lets it go to RG as the subspecies (which it probably is).