Add "often confused with" warnings

Often with species the photo comparison works quite well, and the species are identified the correctly. However, many people who think they see a species sometimes get it wrong because the photos look similar. In person sometimes it’s more obvious.

In general this method of community ID has the one flaw of people agreeing with an identification without solid field experience, or knowing that there are similar species. Once a photo is labeled as research-grade, it goes into the research grade pile of photos, then other people think that is the correct identification, when sometimes it is not, and then the problem snowballs.

For example a small Lupinus nanus can be confused with Lupinus bicolor, and a large Lupinus bicolor can be confused with Lupinus nanus.

I think that in the suggestions there should be a yellow exclamation point Or a red triangle or some sort of symbol that means often confused with, then list which species.

As it stands it’s probably going to take a year or more to go through all of the lupinus bicolor observations and find all of the mis-identifications.

The Jepson manual for example in their treatments of species will have a note about “sometimes confused with.” So it could be something like that.

You may already be aware that on the taxon pages, there is a “similar species” tab, populated I believe by actual ID-change data from iNat observations as community identifiers weigh in.

Currently on the Lupinus nanus taxon page there are six other “similar species” listed (not all lupines!). Lupinus bicolor is definitely listed first though.

But yes, making this information more visible, accessible, in-your-face even, might help slow down some knee-jerk IDs and better educate users.


I really like the idea of using similar species in the suggestions. This could be especially useful if a species has fewer than 20 observations and is not appearing as a computer vision option yet. Perhaps one or two of the suggested options could be pulled from the “similar species” to the top species, instead of from the computer vision suggestions - or perhaps that is just too complex to program?


This would be neat, maybe as part of the proposed user-created ID notes in the taxon page. The problem is these look-alikes are very specific, like maybe the two species only overlap range in one area, or are hard to tell apart with vegetative leaves but easy with flowers. I feel like such a feature would have to be user-created somehow. The devs would never have time


i think this would be best implemented alongside something that allowed people to create checklists on taxon pages that would be tips for identifying those taxons. for example, on the bird (Aves) taxon page, you could have 1. beak, 2. feathers, 3. 2 legs, etc., and on a blue jay taxon page, you could have Aves + 1. blue feathers, 2. crest, etc… the checklists could help to educate people about what to look for and capture, and they could become the basis for a higher-tier of research grade (the more checkboxes can be checked, the potentially better the ID).


This would be fantastic if displayed to user as proposed. Currently the information is on a panel that I suspect very few users consult.


Correct, I never look at the similar species page. I didn’t even think about using it for this purpose. I don’t think most people consult the taxon page.

Perhaps the verbiage is better described as “sometimes confused with.”

Maybe the warning icon could be a link to the similar species page.


I would add that although iNaturalist may suggest there are similar species, I take these with a grain of salt. The computer vision is still much worse than human vision and experience. So, while this site may list more than one species that is similar, the Jepson eFlora treatment for Lupinus bicolor only speaks about L. nanus as a possible look-alike.

1 Like

This discussion actually reveals a larger problem with computer vision suggestions, and the psychology of making an identification. Everyone is encouraged to make identifications, and there are incentives for this.

The problem is people put too much faith in the computer vision algorithm at times, and also too much faith in other people who identify something. This can lead to what I call “bandwagoning” and false “research grade” observations. In a sense, this site values good photos over accuracy, although the community ID gives an impression that identifications will be corrected. Perhaps they will, but they might not.

In general this calls into suspect all observations of certain species. The accuracy of this entire site is in fact, a concern.

I don’t know what exactly needs to happen, but I feel a different way of making an identification research grade needs to be implemented. For the time being, I would take all, Lupinus observations, for example, with a grain of salt, as far as accuracy. It is not screened as well as CalFlora/CalPhotos, although I have seen mistakes there as well.


Just so I’m clear, you’re wanting this symbol next to computer vision suggestions? I’m not sure what you mean by “photo comparison” earlier in the post.

Photo comparison is when a user posts an observation, then decides the species based on comparing their photo with those in the database. As opposed to knowing the traits of a species, such as dichotomous key features to clue in on.

Say I see a Lupine, and I don’t use a key to identify it, then I take a photo. Then I use iNat, browse the photos posted of the species and decide what it is. This is not adequate for some species because they are hard to identify without certain macro shots with a scale or notation by the user that they made certain key observations.

The warning would be next to the species on the list that pops up if you choose the “compare” option with suggestions.

As far as when and where… I’m not sure now. I would hope to keep it looking nice on the observation itself. I suppose it could be by the compare button, once someone picks a species. But, when the list populates for Computer vision suggestions, maybe on that list somehow? It is kind of a chicken and egg problem, now that I think about it.

Right now the only data we have on similarity is from observations that have been identified as both species, which happens a lot. Basically, then, you would probably see this for every suggestion, making it not really any different than it is now. Creating a similarity database that is a bit “smarter” would be possible, but pretty difficult, and then it would have to be explained, also pretty difficult. So it’s not something we’ll be moving forward on in the foreseeable future.

I think a better solution, which is also complex but is something we’re working on, is to train the model on more higher level taxa and provide and emphasize these coarse ID suggestions (like genus and family), as well as remove/alter/makemoredifficulttouse the Agree button.