Recruiting more identifiers

I would like to see the goals you describe met (recruiting more experts, getting more mid-level users to spend more time ID’ing, and making people more aware of the value of their IDing) but I also think there is a missing piece of the puzzle here, which might be more important than all of these goals: helping people to become better at identification.

Identification isn’t something where you become an expert all at once. It’s a lifelong process, and the more you learn, the more you realize how little you know.

“Experts” are often experts specifically because they know more about how little they know, and how many things they actually need to check or exclude to be certain of an ID. Lots of people can identify a dandelion (Taraxacum sp.) and a slightly smaller group of people can identify white snakeroot (Ageratina altissima) in its range. Far fewer people know that there are two dandelion species in much of the continental U.S., the more common Taraxacum officinale and the less-common red-seeded Dandelion, Taraxacum erythrospermum, and even fewer people know the characteristics to tell these apart. Similarly for the species Ageratina aromatica where it overlaps with white snakeroot in range. Even fewer people have a sense of the range of natural variability in things like leaf shape, to where they can see an isolated, atypical-looking plant of one species, and know for certain what it is.

A newbie though might make elementary mistakes. A plant with whorled leaves? Can’t be white snakeroot. Oops, but it is. A plant with opposite leaves? Can’t be Verbesinia alternifolia. Oops, but it is.

Moving from a newbie towards an expert, involves learning how to know just how much you know…those crows you’re observing in the Mid-Atlantic U.S. American crow, or Fish crow? Do you know a chipping sparrow well enough to know when you’re actually looking at a clay-colored sparrow? And perhaps more importantly, do you know when to say “Sparrow sp.”, “Warbler sp.” or “Crow sp.”? It’s like Kenn Kaufman’s book Advanced Birding. I love that book and it made me better at bird ID specifically because it made me realize how I had been IDing stuff that I didn’t truly know how to ID.

It’s about having seen enough plants that you know how to spot atypical variants of a plant, like opposite leaves on a plant that normally has an alternate leaf arrangement, or whorled leaves on a plant that is normally opposite, or a plant with unusually wide or narrow leaves, or a bird with aberrant coloration.

iNaturalist can get better at guiding people in these directions, and we as users can also do the same thing. Some things I’d like to see:

  • Users sharing more detail in comments about how they’re identifying things. “Note the such-and-such characteristics, excluding such-and-such other species,” “No, this is definitely not a such-and-such because it has these features,” or “But that species has never been reported this far east.” etc. This helps educate others about correct ID, as well as other taxa we need to be checking against, and also helps resolve disputes and more quickly reach a consensus. And by mentioning things like range and habitat, we can help other people realize how central range and habitat are in ID, something newbies often under-emphasize.
  • Users asking each other about how they identified certain things: “How did you know this is a such-and-such? How did you exclude this species?” etc. This helps encourage other users to engage in the sort of educational sharing in the previous point, and it also helps resolve disagreements.
  • A UI and feedback system that encourages users to exercise more restraint in identifcation. The system as-is seems to encourage haphazard guessing.
  • Highlighting and more linking to external resources for ID, which can include books, websites, and online communities. This could happen both centrally in the official UI and documentation of the site, and from users in their comments. This can involve citing or mentioning books, and linking to websites. This can include both material on characteristics for ID, as well as range maps, and discussion of habitat.
  • More users setting a positive example of exercising restraint in ID. For example, if you think something is probably a certain species, but you’re not sure, and you are sure of the genus, then identify it to genus level, and say in the comments: “I think this is such-and-such species, but I’m not 100% sure.” This will set a positive example for other users and train them in how to become better at ID. You may also spark some dialogue that helps you improve your ID, if someone can point out characteristics that you might be missing in the photo, that can help you to be 100% sure of the species-level ID that you correctly guessed, or if someone corrects you and it’s actually something else. It also helps if you really have no clue what something is but have a vague guess, like “I have a vague hunch this may be in the Asteraceae family, but I really don’t know.” Learning to distinguish relative degrees of certainty and uncertainty is important for getting better at ID.
  • Sharing, and discussing, aberrant examples of common species, including verbally describing what is aberrant about them, and how we ID’ed them as what we did. This is critically important, both for training people to ID aberrant individuals, and for raising awareness that these things happen in the first place (which can lead to people exercising more restraint in their ID’s by realizing how little they know.)

Basically, the more I think about this, the more I realize that this is something that all of us can do a lot about through our own behavior on the site.

But I also do want the architects and administrators of the site to be thinking about this stuff. I think that the UI of the site may be inadventently encouraging sloppy ID, and at best, it’s not doing all it could be for encouraging people to become better at ID. I’d really like to see an interface that encourages caution and restraint in identification. I.e. we want to make it easiest and most natural for people to exhibit the best behavior and not rely on people to exhibit unusual levels of self-discipline, which I think is what the status quo feels like to me.

11 Likes

nathantaylor this is a complete side-step from your suggestion, but in reading it I thought “I wish I had a better feel for the subgenera of Euphorbia (or Carex or Hypericum or…)” and it occurred to me to check what happens if I enter a subgenus in the iNat “explore” box – whoa! a visual guide to the subgenus! Every day I discover some new amazing thing about iNaturalist.

1 Like

I love your suggestions about what identifiers can do to help people learn. I do try to do this some of the time, and if the recipient displays any interest, I do it more.

I’d like to hear others comment on whether the current system encourages sloppiness. I’m thinking of a colleague who’s a biology instructor but has very little field experience, but wants to learn outdoor IDs. She accepts the first computer suggestion that looks reasonable but always checks back to see how that went, and learns from the correction. Is that sloppy, or just one style of learning?

4 Likes

That sounds like a win-win for any taxon! Sign me up.

1 Like

I guess if anyone is interested in a little Euphorbia training session, send me an iNat message saying that you’re interested, and the discussion can be continued there so as not to distract from the focus of the discussion here. Depending on the amount of interest, I can put together a journal post for easier discussion afterward. @arboretum_amy @janetwright

1 Like

There is another topic with things like this:
https://forum.inaturalist.org/t/help-me-identify-non-experts-welcome/2915

1 Like

I think this would depend on what “see how that went” means.

If someone is just looking at the suggestion and saying: “Hmm, that looks right.” then I’d say that’s sloppy.

But if the person looks at the suggestion and then goes through a more rigorous process of ID, which might involve going up to a taxa (like family or genus) which one knows enough to be sure of, and then looking at the possible taxa that might occur in the range and habitat, and then looking at the plant’s characteristics to eliminate any that might possibly look similar enough (In many cases this could be a really quick process, if the species is very distinctive looking) then I wouldn’t call this sloppy at all.

It’s possible people can make mistakes in this second, more detailed process, but I think they would be few and far between. The huge number of misidentifications that I see on iNaturalist, suggest to me that most people are following something more like the first one, which I think is sloppy, and makes the data lack integrity and be less useful in a long list of ways.

2 Likes

I’m always keen for learning opportunities, especially with plants/areas I am not familiar with but encounter often. Of course it comes down to time, work for me is flat out at the moment, but for the Euphorbias I would find time :)

1 Like

I think any action that is followed up by an assessment of it’s effectiveness (with any appropriate adjustment to the process) is as perfect an approach as one can hope for!

A few of the really great identifiers on BugGuide have already made their way over here and I love to see them working on both sites, though I can only imagine how much work they are putting in. My one worry is that I left BG because it got particularly gatekeeper-y several years ago and I’d really hate to see iNat go down that same road. But unlike BG iNat welcomes all photos not just the best of the litter after a certain point.

3 Likes

You just hit the nail on the head why I stopped using BugGuide - the day I got a message telling me ‘this species is already documented in Ontario therefore your record is of no value, so I deleted it’. Not even the ‘photo frass’ (users who dont know what frass is can look it up) issue, just, ‘you add no value’

2 Likes

Hi all,
It just occurred to me that perhaps folks are right in awarding badges in some manner for helping with identifications. At first I was not in favor of what I see as “ramifying” the process, but I think my mind is changing because I’ve noticed how nice it is to get the recognition on this forum in terms of “Earned Member”, “Earned Reader”, “Earned Enthusiast”. Just a thought that maybe this can be applied to identifiers in some way.

4 Likes

Yes, I sort of like the idea of having “levels” of identifier or something with a little badge, rather than just a count that encourages competition. That way it emphasizes the service aspect.

2 Likes

I am not a qualified identifier (yet) but I went to the Identify page just to see what the experience is like if I WERE an actual identifier. I have to say it would be very frustrating due to the poor image quality of most photos.

If the goal is to get the most species identified in the most places for research purposes, and :

  1. you have limited qualified Identifiers
  2. you have lots of interested users
  3. you want to recruit new identifiers

then I would recommend you make the most of those identifiers that you do have AND find a way to leverage all those interested users who are not yet comfortable doing IDs.

If average users could rate photo quality as Good,Better, Best as a part of the Identify page (without actually ID’ng the subject) then the few identifiers we have could filter on Best or Better images and make better use of their time. Potential new identifiers would find a much more satisfying experience rather than wading through pages of unidentifiable images.

I can certainly distinguish between a clearly usable photo and a likely very poor photo. I’d be happy to help in that manner until I am actually better qualified to do IDs myself. Food for thought.

That may not be as straightforward a proposition as you think. It’s not difficult for a trained observer to distinguish an American robin from an European starling with the help of an underexposed, blurry picture where 3/4 of the bird is covered by vegetation. What constitutes a “good” photograph has little to do with whether it is useful for high-confidence species ID.

The current system is pretty good. You can jump in with both feet and make IDs at the level you’re comfortable. Birds, plants, etc. You’ll be wrong sometimes, and you’ll get corrected by other folks, but you’ll get better with time, and more confident with identifying at a more detailed level.

A lot of what makes iNaturalist good is that novice observers with a cellphone can engage with more experienced users. That engagement would be compromised if you kept the experienced users away from the novices by actively demoting technically poor photographs.

11 Likes

I think the points you make are entirely valid. I am often surprised by what others see in my own photos that I cannot see.

In my suggestion note that Identifiers who ARE willing to view all photos could certainly still do so. Identifiers we so badly need who may NOT be willing to do that could filter the photos and feel like they maximizing the use of their time. Even less skilled identifiers (who may need a near perfect image to make an ID) would benefit.

Sometimes it requires ID knowledge to even be able to tell the “quality” of a photo. An extremely “low-quality” Sciurus photo in Boston is still identifiable – Sciurus carolinensis is the only option. The exact same photo in Texas would be left at genus because it could also be Sciurus niger. You need that identification knowledge to be able to mark the photo identifiable or not.

4 Likes

Any global mechanism that cordons off some users’ observations based on perceived quality runs counter to the philosophy of iNaturalist (as I perceive it).

This doesn’t describe anybody who I’ve had occasion to interact with – online or in person. That’s just my own experience, of course.

1 Like

Lot’s of good ideas here. And I’m glad @janetwright brought it up.

While I share everyone’s goal of getting more people to engage with iNaturalist as identifiers, I want to push back a little on the suggestion that the number of observations is “overwhelming” the system. I think we should all temper our expectations on how much of a difference adding more identifiers is going to make.

Only a few comments have actually included objective data on identification rates–especially @pisum’s great summary of ID status across groups. It shows that the identification rates for most taxa are actually pretty good. 95% of birds identified!!! That’s incredible. I don’t think it’s realistic to imagine that number going up much. In fact, I would wager that that number will go down as more bird observations are added from the tropics where guidebooks are less complete and there are more cryptic species that are not identifiable from photos.

While much lower identification rates for plants, fungi and invertebrates reflect a lack of expert identifiers to some extent, do you really think they’ll ever get up to 95% or even 80% research grade? Two much bigger factors are 1) many of the observations are not identifiable to species, genus (or even family or higher) from the photos submitted and 2) they could be identifiable but keys or descriptions are not widely available.

My response is technology and education. On the technology side, iNaturalist could build algorithms that take into account image quality (focus, lighting, color) and size of the subject in the image and then use these features to determine the probability that the identifications can be further refined. I think an “identifiability” or “image quality” score based on these features could be useful for directing identifiers to the observations that are most likely to be “under-identified” (to coin a term). This system could also give observers some feedback on their observations and help them choose better images to submit. This would just be an assistance system and it could be continually refined and tweaked, I’m not suggesting it count against research grade or anything (so don’t get upset!).

On the side of education, a worthy use of the photos and data from iNaturalist would be to build better field guides and keys and make them accessible for free online to everyone. Many of the taxa on iNaturalist have rarely been photographed in the field and so we are just beginning to understand how we can identify them from photos. This should be an exciting opportunity for mycologists, botanists and entomologists (and others). That said, I’m a little bit puzzled that iNaturalist stopped developing its “guides” feature. If super-identifiers are feeling overwhelmed by the number of observations on here, my suggestion is that they try making some online guides (either on iNaturalist or offsite) to teach the rest of us how they make identifications. That would help a lot.

To paraphrase: give a man a fish identification, he has one fish identified, teach a man to identify fish, he identifies fish for a lifetime!

8 Likes

I totally agree with your philosophy here. I love the democracy of the iNaturalist system and I don’t want to discourage new users by downgrading their observations. But I believe that machine learning algorithms could be trained to judge the “identifiability of images” and that this could be part of the system.

You are certainly correct that blurry aesthetically unappealing images of birds can be identified (just take a look at my observations!). However this is something that a machine learning algorithm could take into account on a taxa by taxa basis. Many photos of plants and insects are just far too zoomed out or out of focus to be identified. Sorting through these can be exhausting and I think it discourages some identifiers which is a bad thing for everyone, including the observers. I’m not suggesting we remove these observations or downgrade them in any way–but it would be nice to have (an optional) filter that could highlight the observations that are most likely to be further refined.