Western bees named with eastern N. American species suggestions/CV over-generalizing?

Please fill out the following sections to the best of your ability, it will help us investigate bugs if we have this information at the outset. Screenshots are especially helpful, so please provide those if you can.

Platform (Android, iOS, Website): Does not matter

App version number, if a mobile app issue (shown under Settings or About): N/A

Browser, if a website issue (Firefox, Chrome, etc) : N/A

URLs (aka web addresses) of any relevant observations or pages: https://www.inaturalist.org/observations?project_id=89326&taxon_id=199024&place_id=any&verifiable=any

https://www.inaturalist.org/observations/123179979

https://www.inaturalist.org/observations/123158176

https://www.inaturalist.org/observations/120429793

Screenshots of what you are seeing (instructions for taking a screenshot on computers and mobile devices: https://www.take-a-screenshot.org/):




Description of problem (please provide a set of steps we can use to replicate the issue, and make as many as you need.):

We have a project to collect all the bees in Washington state. From the page, when we sort by species we discovered that many people are adding eastern North American bees. While range extensions are not impossible, (my Andrena milwaukiensis is a perfect example), I believe that what is happening is this: The eastern bees are fewer and better documented than the western bees. The CV “know” more about the east and it is using them to try to ID our bees out west–consequently we are writing multiple people to check the range before choosing the names.

I tried looking up adding features about range and anything I tried was closed. These suggestions are sometimes out of bound by half a continent. The CV ought to show the genus alone, maybe a subgenus for bees that are that far away. I guess it would be a feature to ask it to say that this is a range extension and to check the comparison carefully…

Nomada in particular are a mess taxonomically and will be unless a taxonomist comes along that wants to do that for their life work. We have bees in our collection named Sp. 1, Sp. 2 through Sp. 12, plus some others with A, B, etc. It is really going to be pretty much impossible to ID them out west to species and get 100 bees for the CV to learn them. It is a conundrum. Having the ranges. limit the suggestions would help.

However Bumble bee ranges are much more defined and we keep getting over-generalizations for those as well. Yellow male Bombus are the worst for getting named the wrong things. We made an Observation Field to sort Pyrobombus, but then need to train people to use it. It would help if something could pop up on the yellow-faced photos to share that.

Step 1:

Step 2:

Step 3:

There’s nothing surprising, that’s the cv.

1 Like

This has been a problem for a long time, but it’s not exactly a “bug”. Computer vision suggestions don’t take geography into account (although lately, visually similar taxa that aren’t “Seen Nearby” are hidden by default). As of April, they say they’re working on a solution:

Future work
In addition to training and deploying this new model, we’re working on a few areas to generally improve the quality of suggestions.
First, we are continuing to work on new approaches to improve suggestions by combining visual similarity and geographic nearness. We can’t share anything concrete right now, but we’re excited to be working with some Visipedia researchers on this.

Until there is a more fundamental solution, more observations and identifications will help get Western species into the CV model.

2 Likes

Yeah, it’s not a bug, it’s how the CV model works.

For what it’s worth, I checked out https://www.inaturalist.org/observations/120429793. If I click on the ID field I see this:

If I click on “Include suggestions not seen nearby” I see this:

I wonder if the observer doesn’t have the “Only view nearby suggestions” toggled on.

For an observaiton like that, if you think it can’t be IDed to species, please add an ID of Nomada and choose the orange button when you get the Potential disagreement pop-up.

1 Like

In examples like people IDing Apis cerana for mellifera in the US, species seem to be automatically added to the Place checklist once one ID is made, even if not RG. This then causes the species to be added to the checklist and shown when using the Suggestions/Compare checklist filter. I wonder if this also results in it being shown in Nearby, given that it’s in the checklist. If this does occur even before RG, one solution would be for iNat not to automatically add taxa to checklists until they’re RG.

No, nearby doesn’t use the checklist, it uses actual observations.

2 Likes

Okay, although the wrong species get added to the checklist filter, which can perpetuate misidentifications for any who use that filter. Requiring RG before adding species to any list would seem best.

Thanks all for the useful information. Also we find so many people just follow the leader, if they get one RG they think their Bee is identified and happily agree. I have a journal post to tell them to please check the range and if it is a far eastern bee or not, but some folks take many wrong IDs (with notes from me explaining please don’t) to learn this. Also we have extended several ranges with proper IDs, so we can’t go overboard overcorrecting either.

1 Like

You mean species id, there can’t be one RG to agree with, Research Grade is just a status of observation.

1 Like