I have the impression that in recent months (3 months? 9 months? I’m not sure) it is taking longer for even very easily identified animals in areas with lots of iNat users to receive identifications. I say this based on the feeling that fewer of my observations are getting IDs in the first days and weeks, but also based on what needs ID when I go to identify. I mostly identify vertebrates and the more common invertebrates in my county, and previous to this year I hardly ever saw the observations of the really easy birds (Western Bluebird, Great Blue Heron, etc.) beyond the first hours because others would ID them before I got to it. Now when I identify, there are unIDed observations of these things (not just in birds) that are days or weeks old.
My question is less “why is this happening?” and more “is this just my imagination?” Are others noticing something similar? Thank you.
Looking at the stats page https://www.inaturalist.org/stats the number of observations being submitted has gone way up, but the number of observations being identified by the community has remained relatively constant. I think that does imply a longer time for an ID to be made on a given observation.
As always, iNat puts more emphasis of getting new users to submit things than recruiting people to identify (why is there no identify page the app???!!!??). i digress.
I’m not surprised. I do a lot of identifying, but I have noticed that the rate of new observations is accelerating to the extent that it feels out of control.
I also have things I would like identified, but I know that once observations disappear off the first 10, or 20, or 50 pages of “Needs ID” they aren’t likely to get an ID except by chance sometime in the future.
I agree this is happening, and I agree that it’s driven by the influx of new people learning about nature versus specialists managing species-level IDs.
I suspect that in the long term, iNat may have to change its incentives. I’m a generalist / local specialist and I try hard to give equal effort to observing and IDing. I make mistakes and even though I know mistakes are the price of involvement they’re embarrassing. I’m also less inclined to say e.g. “Yes this is definitely a boneset but I don’t know species” when someone has already made a species guess, even if it’s from c.v. Therefore I think encouraging lower-specificity confirming IDs would probably be helpful both for scalability and pedagogical purposes.
I’ve toyed around with some ideas for tracking how well we produce IDs at multiple scales of precision. For example, I like the idea of saying an observation is “research grade” at the family or genus level even if the species is still being debated and refined. I’m also interested in how many of my IDs are accurate to a given level of precision and making that more visible: If I ID “sensitive pea” for example, I might be wrong 25% of the time, but maybe I’m right that it’s in the genus 80% of the time, with a few major errors like Persian Silk Tree saplings. It feels like nuance here could help scalability as we have more amateurs / generalists relative to specialists / experts.
Perhaps there are also some historical large-volume identifiers who now contribute less. I am one of those.
The last City Nature Challenge produced an avalanche of very poor quality observations, with lots of cheating. Not so much in my own area, but in certain CNC hotspots where I tried to help out. That was a tipping point for me. It became impossible to keep up and quite disincentivizing. The result was that I longer attempt to provide identifications for all observations of my specialist group in my area. Now I focus on correcting misidentified RG observations which are generally the result of the CV/Geomodel (both with poor results for my group) in combination with blind agreements. That requires much less time and keeps the overall data quality reasonable. It means I provide less help to those trying to learn, but hopefully others will step in to fill the gaps - before they too get burnt out.
I was thinking the same thing just recently. I also get the impression it started around the CNC, which would make sense. I’ve been trying to ID more (especially some of the harder stuff that I used to never touch, since that’s what really gets stuck in my experience) but I’m quite slow as I try to be very cautious since I know I’m lacking expertise.
By the way, if we look deeper to the stats, we can see decline of active users count, from 490+k on May to 395k now. There was also some strange moment gap from 488 to 430 in the beginning of the June (may be some spam cleaning?).
Also there is traditional summer declining when many active users, especially scientists, go to the fields for expeditions, and begin to active upload they observations. May be it will be a bit better in winter?
PS: By the way, this month Inaturalist can reach 300 millions of verifiable observations. It is 296.5M for 20 of August.
As someone with 20x more IDs than observations, I can relate to this. The last CNC left me with some measure of feeling that there’s no point.
I don’t think this is something that’s contributed to the problem for me, though, as there was no CNC in my area. I used to be able to keep up with most of the IDs of observations in my group and in my area, but I can’t keep up, not even close. Especially the more that iNat is used for education purposes. Student users in my area are generally being given so little guidance on how to use iNat that, I’m sorry to say, that the quality of their observations has been problematic.
I’ve tried talking to people who are obviously getting students to use it, and encouraging them to spend more effort instructing them on what it is, and what it isn’t.
I think there increasing usage of iNat just to get an ID on an observation at the moment it is made, with an increasing proportion of those users never checking back into their observations. In this way I think Computer Vision is working as more of an impediment, detracting from the amount of effort that duress or new users put in.
I ‘wasted’ about 2 months on the Cochabamba CNC mess - and still have notifications trickling in from brave identifiers who persevere. With hindsight, I wish I had used that iNat time far better.
iNat needs
better onboarding to improve the quality of obs, so they can be IDed
reminders to respond to Notifications (but Notification management is a huge problem)
to encourage observers to pay it forward with IDs. Or Annotations. Something at least.
generally to encourage IDs. Waiting for better management of Notifications is a barrier that has put off identifiers.
I admit defeat. Cannot keep up with my chosen slice. Have tweaked my bookmarked URLs down to what I can clear daily. Chewing away at my backlog, where I can enjoy adding value. Abandoning great swathes as Reviewed, Next!
PS note to self. Remember to avoid this upcoming project in October!! (Their CNC project is Suspended - but the broken IDs live on, forever)
In the northern hemisphere at least, the fact that the last three months are probably the best months for being out-and-about observing could be a significant factor. I know I’ll be spending much more time in front of the screen IDing from October/November onwards.
So am I. In July I’m able to make as many observation at one day as in whole February. And I need time to convert photos and upload the observations to the site. Today I work on July 3, and yes, the observations, especially if I didn’t identify to species level, could be stay in unidentified status for weeks and months.
Yes. Identifications are taking longer. The obvious evidence for me is that I can find easily identified North American birds still unidentified eight or ten pages in when I’m identifying random current observations, but this is my general impression also. Why is it happening?
Season, for one thing. Many of us in the Northern Hemisphere have been out in the field and then processing our own observations so less time to ID.
That last CNC was discouraging. I’d helped a bit with the Bolivian CNC’s the last couple years (annotating lots of cultivated plants but building up a very short list of native plants I could ID) but this year it just felt overwhelming. It didn’t turn me off from all IDing but I gave up on that.
Notification management is deeply frustrating. It’s hard to get back to work on IDing when I have minimized time on iNaturalist for a few days and see I have 567 notifications and know that’s just the tip of the iceberg since the number doesn’t include the observations from people I follow. Why is that a problem? Because I can’t just ID the first few pages and later return to it because I’ll have to page through those first few pages again to get to the ones I haven’t ID’d. I have found a work-around – I quickly use the “more” button to go in 20 to 40 pages and start there. Of course, if I misjudge and can’t get them all done in the time I planned to I just have to keep working because if I stop I’ll have to go through the whole process again. I’ve gotten to bed late some days because of that.
It doesn’t help that the order of notifications varies somewhat each time I open my notifications. Sometimes I see something I know I worked on last time and think, “I’m done!” but then realize I haven’t see the next few. So I always have to go a page or two beyond the “end” to be sure I got there. All this wastes time I could have spend IDing.
Tony pointed out that we can change settings to reduce the number of notifications we get, but it didn’t help as much as I could wish. I could avoid seeing ID’s that agree with mine, but (1) I like to see that I was (probably) right and (2) I can just scroll past them quickly, just opening an observation if I have a question about it, so they aren’t a big problem. The thing is, I follow some prolific observers (though I no longer ID all the observations for all of them – sorry) and I have gotten so I often add the observations to projects, which takes time, too. Plus, of course, the more I ID the more notifications I get . . . .
We really do need more people identifying. I’m not sure how to recruit them.
“We really do need more people identifying. I’m not sure how to recruit them.”
I’m giving an iNat training to our new naturalist class in a few weeks. I plan to reverse the usual order: to start with how to use iNat to see what species they can look for, then talk about how identifying observations can be a great way of learning, and about best practices there. Then, once we’ve identified some things to our level of ability, talk about what it takes to make a good observation, and how we can consider making our own. I’m optimistic this will help the class avoid a lot of typical new-user errors, and get them in the habit of IDing and annotating early.
Those JRMN usually post some great images of observations in the park. I rarely have issues with the quality of their photos. The majority of them still automatically select the CV suggestion, but that’s expected of most new users. As long as they’re not confirming each other’s IDs just because they’re friends, I think we’re doing alright. The only time I see them post questionable photos (and this is the case for all master naturalists across the continent) is when we do the aquatic macro trainings and everyone is posting grainy, underexposed images of dark squiggly things in the ice cube trays. We need a better system for that.
I agree that I’ve seen a drop in IDs on my observations, including things that were previously IDed fairly quickly. My expectations for getting IDs have definitely dropped. Personally, I’ve done less IDing this past year. I think the two major reasons are that, while I used to try to keep up with certain taxa, this has become clearly impossible, and this year’s CNC filled me with a sense of futility. So now I ID when I can/want to and figure that is enough as opposed to trying to keep up.
