Yes, that is a good key. Any way we can make it more efficient and effective for someone to become an expert is good. We really could have a system of good graphics and annotations for each region available to conservationists, so that one could become an expert within a month or so. That is where we need to go.
What if we could add/remove a single unit of similar species per species in the similar species list.
For example the many-line lucine (which I have hundreds of) only has one similar species listed (dosinia-like lucine). Letâs pretend the dosinia-like-lucine looks nothing like the many-line lucine (in reality the juvenile looks very similar but Iâm just to lazy to find another example) because misidentifications on completely different shells do happen all the time (the CV is really bad at sea shells) and are completely useless if someone is trying to figure out if thereâs a look-alike by using the similar species list. The feature would allow me to remove one unit of Dosinia-like-lucine misidentifications, and since thereâs only one misID, it would disappear completely. It would also allow me to add in the lovely miniature lucine and Clathrolucina costa which are the exact same sizes shells and could be easily confused by someone unfamiliar with lucines.
When I first started on iNat, I relied heavily on the similar species list to know if there was something my organism could be misidentified as, and my first couple hundred IDs all depended on it too. If all the similar species looked different/did not live in that area, Iâd assume that it must be that species. I donât do that now, but I know that a lot of new users do.
What do youâall think? I know itâs not at all the extent of ID information yâall have been talking about, but itâs seems to me (who knows little about coding) much easier to implement than a complete addition from a respected identification website. If that did happen, I would want this website for Floridaâs marine gastropod and bivalve species: https://olram9.wixsite.com/letstalkseashells/copy-of-presentations-univalves
because it has lots of pictures and emphasizes the difference between similar species.
1 Like
Seriously, this needs to be implemented already! There are tons of identification tips that are buried in comments that hardly anyone will ever see again. Links to other sites are not helpful because general users are not going to A) know to click on them to get ID tips, B) 90% of external links will not have identification tips, so knowing which links to follow to find identification tips will be nigh on impossible, C) external sites may not include all the relevant tips or not include them all in one place - donât make users hunt around for the info
As recommended elsewhere, keep it simple, it doesnât need complicated voting mechanisms, or linking to tips in comments, etc. Just make it a wiki-style page that will handle everything from simple bullet lists on up to well formatted pages and images (including links to other sites or example observationsâŚ). Have a âTalkâ tab so users can discuss what to include or not include.
Is this replicating Wikipedia? Yes and no⌠it may be replicating the functionality, but Wikipedia articles rarely include tips on identifying a specific species. So the Wikipedia article is rarely ever going to include all the tips that someone would want to put on this iNat identification tips page.
Donât worry that tips in some regions of a species range may not work in other regions â that can be sorted out in the page if users can add sections specific to each region.
Just start with something simple (but robust) and let people start doing this. If new features are deemed useful down the line (i.e. voting - for some reasonâŚ) then add it later once it becomes apparent that it would be helpful. But the ability to simply list some key features to help identify a species would be a huge win for iNat!
7 Likes
Actually, my idea would totally screw up how the math worked in the similar species tab. Itâs not worth considering. A better suggestion would be to be able to add species to the list but for them to appear at the bottom with a 0 at the top right signifying there have not actually been any misIDs yet. I wouldnât think that would mess with the numbers of the automatic system but nonetheless allow you to add species. (I actually made a feature request but it was denied as editing the similar species list was stated in 2019 as not going to be done). Since those topics closed many years ago, the only place I can comment is here where it is slightly off-topic (upgrading the tab vs making a new one)
If only I could missID one or two to make them show up in the similar species list (to be immediately resolved and at the observerâs consent of course) but that is grounds for suspension.
1 Like
I think this is a good idea. It could also be used to include confusion species that donât have observations on iNat yet, alerting observers and identifiers to the existence of these species.
1 Like
If there was a way to bottle that âtribal knowledgeâ and make that available as an enhanced identification guide, that would be amazing.
Yeah this is my #1 dream for iNaturalist. The other day, my friend had someone tell her that nudibranch species A was actually nudibranch species B; the only reason the identifier knew that was they were up on very very recent literature describing a minor but visible on typical photos difference in appearance between those two species.
The other thing is for most less-characterized (non-bird) taxa, thereâs a lot of just⌠spurious ID tips out there. If I read on some identification key website, university extension, or what have you, that you can distinguish species A and B by characteristic â1â, but thatâs actually not true, or only true for one sex of the species, or only true in one specific area but in the area Iâm looking you need to watch out for species C as well.
Thereâs a ton of people on here in the trenches every day fixing mistakes by posters and IDers, but it seems Sisyphean at best.
At first I was like, just add the distinguishing characteristics of this species to the wikipedia page so that when people click the taxon description and the wikipedia article pops up (if there is one), they can read that they need to check characteristic â1â and watch out for species âCâ in region âXâ before making the ID to species âA.â But actually thatâs against Wikipedia editorial guidance; Wikipedia is Not a Guide. Okay⌠thenâŚ
I looked through the forums and it seems this issue goes around and around without anything permanent being implemented. There were guide pages briefly debuted (and still exist, but not with any ongoing bugfixes etc) that incorporate local knowledge and dichotomous-key-style questions. Now the official guidance is to post identification guides on your Journal pages, and many wonderful IDers have done this. But again, itâs kind of buried knowledge.
Already, I think the taxon page is trying to take into account where the observation was made (the invasive species pop-up, for one). My dream form of iNaturalist would have on each taxon page, an iNaturalist wiki above the wikipedia article. It would be titled âHow to tell it apart from similar speciesâ or âTips for IDâ or similar. You would be able to toggle this box between âGlobal - include all tipsâ and âLocal - include tips that apply to my localityâ. In previous discussions of such a feature, devs wanted to know where the info would come from. I think doing it wikipedia style would be the best wayâpeople would post tips with citations (e.g. to guidebooks, DNA-verified specimens, taxonomic papers, etc.) and editors/curators of each taxon would be responsible for vetting the page. Each piece of information would need to be contextualized as to where it applies (Globally or in specific regions; if so, which regions) so that the views could be toggled on and off to be the most useful for ID.
Alternately, you could have a not quite wiki style where people upvote the most useful tips, similar to the Community Notes feature on Twitter.
I know this system would be extremely complex to program and to implement, so I realize Iâm asking for the moon. Still, itâs frustrating to go on other sites and they all feature ID tips and then iNaturalist itâs just vibes-based on pictures, and then people who are onboarding get confused and frustrated by simple issues like:
- CV is wrong sometimes.
- The right taxon might not be on the CV suggestion list at all.
- The closest-looking photo is not always the right taxon.
- Things youâre pretty familiar with in your region might look different, or have lookalikes in other regions.
- Being stuck on where to turn for IDing resources (I know many taxon pages have links to good websites, but these are a very very incomplete mosaic and some contain outdated or very limited information).
- No one is identifying their observation after months and moths (they donât know that thereâs no way to identify x down to species level in their region without information y, and experienced IDers are tired of pasting that spiel on every observation in the region.
- People donât feel confident to start or continue IDing because they donât know where to locate information that is very hidden (like my example about the nudibranch).
All of those issues would be addressed at least somewhat. In any case, experienced IDers could consistently point people to taxon page for x region, and a lot of people would point themselves there!
Iâve thought of posting a feature request on this subject, but Iâm not sure if itâs the right move given that this topic seems to have already been discussed a lot with no real progress roadmap. It just gets bogged down in how complicated and difficult it would be to implement, sometimes with the angle that we need to figure out/write the ID guides before the roll-out⌠instead I think just adding the feature and allowing people to add location-specific tips would be amazing.
I dunno, am I being too whiny? I know perfect taxonomic ID isnât the main point
of iNaturalist, and itâs not advertised as an identification tool per se, but the fact is, identifying stuff is the main way people interact with iNaturalist, whether their own or othersâ observations, and so itâs nuts to me that other than the interpretation-free âmisidentified speciesâ tab on the taxon page, thereâs no centralized repository of identification tips; itâs all 100% reliant on very random and inconsistent community interactions. Donât get me wrong, I treasure community interactions that give me a great ID tip, but itâs a shame how having to comment the tip makes that knowledge very ephemeral. What if that person deletes their account? It also makes for a very frustrating landscape for the experts or overseers of some project who have to keep posting the same tips over and over. If only there were some centralized place to put all those tips!
7 Likes
Thereâs been some discussion of this type of idea and feature requests on the forum previously. The most relevant is probably:
https://forum.inaturalist.org/t/expand-the-similar-species-tab-into-an-editable-identification-guide/13890
But this is pretty far afield of the original topic here, so if any additional discussion of this is desired, I can move the post above and any other to that feature request (or any other thread desired).
1 Like
My apologies, I should not have posted so many off-topic paragraphs. The wish for better ID resources was a bit of a rant trigger. :)
Yes, that thread you linked to was one I had in mind. Sure, Iâd appreciate it if you moved it to a feature request thread if people are interested in talking about it again. That was sort of my intention anyway; if there were replies that people were interested in getting back on this merry-go-round of discussion then I planned to start up another feature request thread. I appreciate your proactive response in preventing this thread from getting utterly derailed.
1 Like
I dunno, am I being too whiny?
This doesnât sound whiny to me at all - I think youâve voiced something a lot of us would love to have. I donât think it needs to be anything complex or radically innovative, just somewhere sticky on the taxon page where people can curate knowledge of the current State Of The Art for identification. If not as a subsection of the âaboutâ tab, then on a separate tab there.
1 Like
Thanks. Yeah, I am wondering if previous discussions got bogged down in chasing perfectionâworries about formatting, photos, different needs for different taxa and so on. But literally a barebones box on the taxon page where people could post the comments they keep leaving on the relevant taxa would be a major step up.
@cthawley, can you move my super long comment into a new thread? I do not know how to do that (or if I have the power to).
1 Like
Moved to the existing feature request as a lot of the convo here addresses similar ideas.
You made a really good point about how wikipedia is explicitly not the place to do this - aside from the fact that youâd be constantly co-editing the page with people having possibly very different goals than inat users, thereâs a genuine conflict in what the two groups of users would want to see there if we used it for that.
Something as simple as the ânotesâ section on observation pages for the taxon page would be a great start. Ideally it wants to be revision controlled, so people can see the history of changes to it, but if we can link to references, photos, observations (and the discussions in them), that gives us a visible place to summarise and share the current consensus on useful keys for inat users who only have what the observer could report here to go by.
Until weâre all carrying âphonesâ that do DNA analysis on whatever we wave them near, thereâs always going to be uncertainty over species ID (and there probably still will be even then!), but developing good consensus on what things probably are, and how certain we can probably be about that definitely improves inat for all users of it.
Weâve had some of the comment discussions here propagate into descriptions in Fishes of Australia because its editor is active here - but weâve also had discussions about âinformalâ keys - things which have been noted as visible in images that seem to be âreliablyâ good indicators, even if they arenât part of the formal description of that species - which are useful for building consensus on observations here, but not necessarily appropriate to propagate as âkeysâ into other publications without more formal analysis.
@tiwane thereâs hints of a plan for this from 2019/2020 above - and I donât want to nag, since I know all too well the time vs priority problem, but is there some detail on that? Itâs clear lots of people want this, but is it useful for us to be hashing out what we want from it, or is there already a good plan for that which just needs someone with time to implement it?
3 Likes
You can absolutely add this type of information to Wikipedia and I have done it thousands of times with zero negative feedback. It is in the phrasing.
Just donât phrase the section like this:
âLook out for the purple spots under the wings to identify this organism compared to its lookalike Species C.â
Phrase it like this instead:
âIt has purple spots under the wings, distinguishing it from the similar Species C which has orange spots. [citation here]â
5 Likes
Thanks for the tip about wikipedia! Thatâs encouraging. Have you had issues with pushback if the information gets too granular? For example, listing different lookalikes in different locations or different fieldmarks that are different in different sexes/populations/subspecies etc.?
I think adding more of those will definitely be an improvement and help it make more sense why iNat links to wikipedia as the primary description, but I worry that for longer articles the ID info will be totally buried. Sadly a lot of taxa have stub or nonexistent articles, so adding the ID info will be right there near the top of the page. But for articles like âRed-tailed Hawkâ they are already so lengthy, and the description section already so long, that itâs essentially useless to an iNat user trying to make an ID. It does mention the belly band, a key mark that often distinguishes it from lookalikes in its range, but you have to wade through a ton of info to get there, and even then, itâs not very clear how to use it diagnostically or if itâs even unique to the species.
This is why I would still advocate for moving an identification box to the top of the taxon page above the wikipedia description, if we ever get a system for that rolling. It could even be based on subsetting from the wikipedia article, but my worry is that people (including myself) have the attention spans of goldfish and are used to wikipedia not having a lot of identifying info, so having the info called out in a separate box at the top of the page would go a long way towards people 1) seeing it 2) actually reading it 3) using it.
1 Like
Something as simple as the ânotesâ section on observation pages for the taxon page would be a great start. Ideally it wants to be revision controlled, so people can see the history of changes to it, but if we can link to references, photos, observations (and the discussions in them), that gives us a visible place to summarise and share the current consensus on useful keys for inat users who only have what the observer could report here to go by.
Agree agree agree!!! Yes! I think the revision history might be a bit cluttered/confusing if on the main page, but like wikipedia I think you should be able to click and view the edit history.
Until weâre all carrying âphonesâ that do DNA analysis on whatever we wave them near, thereâs always going to be uncertainty over species ID
I dream of this Star Trek future. DNA extraction and sequencing pipelines have come a long way in the last 50 years but it still involves a lot of drudgery and toxic chemicals.
(and there probably still will be even then!),
Yup, lots of debates in taxonomy about this exact subject (well, not exact, but how to fit the new bounty of DNA info into things).
âinformalâ keys - things which have been noted as visible in images that seem to be âreliablyâ good indicators, even if they arenât part of the formal description of that species
Yessss this is exactly the type of stuff I want added/boosted to the top of each taxonâs page. One cool thing about iNat is it gives a massive dataset of what it looks like to recognize that taxon âin situâ the vast majority of the time, while the species description is certainly made ex situ with things like exact measurements that often arenât really needed to tell A from B from a photo. (Iâm glad species descriptions are hyper-exact and wouldnât change that, but anyway I definitely agree about the different use cases). Even the basic ID tools like using a dichotomous key are often not useful for plants posted on iNat, and so it sucks when the available resources arenât suited to the data at hand. (There are great discussions ongoing about trying to get more tips about how to photograph/record certain taxa to make a fine-grained ID most likely, but thatâs a tangential issue since even if these âhow to record an x observationâ roll out, it will most likely affect only the most serious/long-term users, with new people posting quick phone snapshots every day.)
is it useful for us to be hashing out what we want from it, or is there already a good plan for that which just needs someone with time to implement it?
Yeah my question exactly. Considering that when I log in to iNat it asks me to donate and also if Iâm interested in high-level job openings there, there could be some current barriers to developing and rolling out major changes at this exact moment. But obviously, I have no idea what the day-to-day workload is like or their internal roadmap for the site/app and it would be pointless for me to speculate.
I fully agree with you here. Ever since I started with iNaturalist and getting into identification, I have been wishing for a website that would funnel all guides together and throw the right one at me when I am looking at a certain living being.
It had been really rough in the beginning finding all those seperate websites that might have information for different taxa (and even more is hidden in expensive literature).
I think nowadays iNaturalist is having less of an issue with people taking pictures but more with the quality of the observations and the burden on the relatively small group of people IDing.
I donât think people want to make bad observations but oftentimes simply donât know what things are important if you wanna determine a species by photos.
So the way I always hoped iNaturalist one day will work, is that you take a picture (or point the AI cam at something) and the app shows you the most likely species (or higher taxon). It should than actively let you know that there are very similar species.
Next the app could give hints about what things to take pictures of to make something distinguishable (âtake a sharp photo of the wingâ s veinsâ) or let people know that this animal canât be identified to species level by pictures alone (lots of insects need their genetalia be looked at under a microscope).
The most futuristic thing (that also was somewhat mentioned in the recent image to text blog https://www.inaturalist.org/blog/95911-search-inaturalist-photos-with-text) would be to have iNaturalist explain to you why it thinks thatâs this species and maybe give you photos of that feature from similar species for comparison.
Or maybe even highlight areas that should be focused on.
2 Likes
like wikipedia I think you should be able to click and view the edit history.
Yes, thatâs the model I had in mind - and I just learned (from another discussion here) that the taxon pages already have history tracking (since 2022) - anyone can view it via the âCurationâ drop down on taxon pages. So most of the tools are already in place, we just need a field on or linked from that page to put it in.
One cool thing about iNat is it gives a massive dataset of what it looks like to recognize that taxon âin situâ the vast majority of the time, while the species description is certainly made ex situ with things like exact measurements that often arenât really needed to tell A from B from a photo.
In the case of marine critters, a huge proportion of species have been described by people who have never even seen one alive. New species were Discovered from what was pulled to the surface in trawls, or had washed up on beaches, and the state-of-the-art method of surveying for âcrypticâ species was for a while to dump rotenone into the water then collect whatever floats up.
Which means all of the things that tend to be most evident in photographs of healthy living creatures, like their colouring while still alive, and the habitat they are associated with, and behaviours, donât even get a mention in older descriptions because they could only be imagined. The formal descriptions are mostly of things that donât degrade in a jar of preservative.
So your point about having keys that are suited to the data theyâre being applied to is a very important one.
3 Likes
have iNaturalist explain to you why it thinks thatâs this species
Thatâs not really how the computer vision works. It doesnât break down the image the way a person or descriptive key would and go âitâs got wings with feathers, if it has 2 legs itâs probably a birdâ âŚ
It doesnât know about wings or feathers or legs as such. It just knows that the image you asked it to match is more similar to the thousands of pictures it was shown of species A than to the thousands of pictures it was shown of all the other candidate species.
And youâve got more hope of having a chat about the stock market with a dolphin than making sense of exactly what it considered most significant in deciding similarity. All it really knows is what images it got right and wrong when it was being trained, nobody told it why it was right or wrong, so it makes up its own very machine-centric explanations for that.
What youâre asking for is more like what preceded the second coming of âAIâ as fad word that all the Cool Kids wanted to be associated with ⌠an âexpert systemâ where the computer walks a person through the logical steps needed to make a decision, and shows them what options remain after each step. Which is something inat could do, but it still needs to start with the low level task of having real human experts define the method to use for each and every species group.
1 Like
But that is something they wanna achieve with the new image to text algorithm they are testing. See
https://www.inaturalist.org/blog/95911-search-inaturalist-photos-with-text
Image segmentation has also been around for quite a while. So maschine learning could quite well work on figuring out what a wing and what a thorax is.
Never said Maschine learning algorithms are supposed to write the guides but that the apps would be able to utilize the guides to help people along. I know thatâs probably still some time off but having integrated guides with comparing images neatly integrated into the iNaturalist interface would be a good first step.
Right now guides are pretty hidden and not linked to the species (and location) so you have to be on the active lookout to find the one you might be needing.
2 Likes
Just a reminder that the current Guides feature is not planned for any additional development. Current Feature Requests are not being accepted for Guides (https://forum.inaturalist.org/t/about-the-feature-requests-category-please-read-before-posting/69), so conversation here should focus on different functionalities that donât include existing Guides.
1 Like