Here is range for M. macilenta, the native species which should probably be being applied (I say probably because there are some undescribed species in the southwest that it could be, but I’m splitting hairs and M. macilenta is the best name for the moment.)
M. macilenta is only showing up as the third suggestion on the CV list.
I’ve been trying so hard to get this genus cleaned up in the states, but seeing that it is still suggesting things at the very top of the suggestion list that are way out of range is.. disheartening. I feel like this is worth bringing to y’all’s attention.
EDIT: My GUESS for what is happening here is that the OK obs might be JUST out of the geomodel range for the species, which is causing the odd suggestions. It just baffles me that it is defaulting to what is WILDLY out of range instead of the closest species
As far as I know if you choose “Visually Similar” on the Identify page, it ignores location for the list order. Usually it should say “Visually Similar” beside the “Select” button, as well as “Expected Nearby” if they’re nearby in the geomodel.
In this example only 2 of the 10 suggested species are “expected nearby”:
If that’s the case, I feel like that’s something that should be changed. Suggesting visually similar things that are out of range is totally fine by me, but IMHO the closest things should shuffle to the top (and in my experience, they usually do shuffle to the top.
For example, here is one of my Macrolepiota observations; M. macilenta is being put at top even though another that is out of range that is also being suggested
I agree with you, and it does seem like this would be a change to how the web interface currently works (on the app, visually similar is filtered to expected nearby by default). It would make a good feature request.
These are suggestions in Identify, which uses somewhat different criteria for choosing suggestions than the default interface on observation pages/upload pages, which is what most people adding IDs to their own observation are going to be using.
If you open the observation and click on the “suggest an ID” box, is the CV still suggesting species that are not even found on the same continent?
That box is suggesting things that are… well I’m not going to say PROPER but they’re at least species present on the continent.
The thing is, they’re not Macrolepiota, and the initial ID was a CV Macrolepiota rhodosperma ID, which leads me to believe the Identify page must have been being used.
Or the observer may have clicked on the option “Include suggestions not expected nearby”.
One of the things about bad CV suggestions is that they are not exclusively the fault of the CV – there is also a human element. I fully support trying to find ways to get the CV to make better suggestions, but I think it’s also important to be aware that (barring only giving users a single option to choose from) we can’t prevent users from making poor choices when deciding how to use the suggestions.
Did you test it to see if thats the result you get when you click ‘include suggestions not seen nearby’?
If not, give me a few minutes and I’ll go test it
EDIT: There is no ‘include suggestions not seen nearby’ option on the page that I screenshottd above.
Anyway, I’m fully aware that there is a human element, and you can’t eliminate people zooming past the best suggestions and clicking on something completely random or out of range. Nothing I’ve said previously says that I’m not aware of this.
This post is entirely about the fact that the first option being offered in this case is one that is incredibly unlikely to be correct, and let’s be real, lots of people are just going to click on the first one suggested - if the CV could be tweaked slightly so even if it’s out of the geomodel range, it’s defaulting to putting the geographically closest species to the top, I’m confident that would cut down on at least a few initial misidentifications.
There is? In small print underneath the list of suggestions in the screenshot.
I didn’t check this particular case, but my experience is that in most cases when the observation is not too old and the observer has added a wildly out of range ID, I am able to reproduce their process by changing the setting from the default on the observation page (expected nearby) to include suggestions not expected nearby.
And the second one of my own that I posted earlier, but on the first page. This second one, especially, seems to be the same list as the ‘visually similar’ source on the compare page - it’s not for my first observation because I failed to remove the section filter (I can correct that screenshot if anyone cares that much.) But to me, it appears that when you click to ignore range it is defaulting to the list order on the compare page - which honestly even further supports my point that for the CV, if all observations of that genus (visual similarity) are out of range, it should still be weighting geographically closer ones to the top.
I thought you were trying to determine why observers were entering implausible IDs.
My point above was that the suggestions you get when using Identify are unlikely to be what observers are seeing, because suggestions in Identify do not entirely correspond with the suggestions on the observation page or upload screen (Identify uses slightly different selection criteria).
So implausible suggestions in Identify, while not ideal, are a red herring for understanding the source of wrong CV suggestions selected by observers.
That’s fine, and I appreciate your contribution to the conversation, but I’m not ignorant of how the CV works, and those weren’t really questions.
The fact remains that an observations that is only very slightly out of range for the geomodel for a species is having things that are thousands of miles away being bumped to the top of the suggestion list, and that is a problem that could be resolved.
EDIT: I also want to make it clear that i’m not picking on the user who posted that observations, at all. I’m just talking about technical features.
