User actions with "top suggestions"

[DianaStuder]

Wrong is fine - so long as you respond to notifications.
And if it has sat in Unknown for years, wrong that triggers a response, is ultimately good.

[sedgequeen]

One of the reasons people use iNaturalist is go get names to put on their photos. Therefore, the CV is useful. It’s very often not used with appropriate caution, but so it goes. I really don’t think putting color codes or dots on the possibilities helps much since most people who don’t really evaluate the choices will simply go with the top suggestion (the one with the most dots or the greenest color) anyway. I don’t think this problem can be solved.

[upupa-epops]

@sbushes

But even better would be if there were core iconic taxa buttons to choose from…or ideally, the broader most likely taxon based on the autosuggests ( in this case, Aves ). This feature request addresses this issue and is under-review :
https://forum.inaturalist.org/t/in-cases-where-the-cv-is-not-pretty-sure-of-anything-offer-a-suggestion-of-a-higher-taxa/14781

I was going to add this request here as well, I’m glad you brought it up. Like for example with this observation:

@sessilefielder

In your case, you might find it useful to know that the model is not particularly confident about any of its suggestions for a given observation, e.g. https://www.inaturalist.org/observations/147931454

Sponges and limpets and anemones are all very spread apart in relation to each other, but they are all animals, and an “Animals” ID would be more useful than choosing any one of those species or than leaving the observation as “Unknown”.

By using the lowest common ancestor of the suggestions like this, it would be pretty easy to have a “pretty sure” suggestion for nearly every observation.

@gcwarbler

I think the general user community should only be exposed to the “pretty sure” IDs.

[sbushes]

@taylorse

The alternative is to leave everything you don’t know about in “Unknown”

The alternative (if there is nothing suggested as "pretty sure of ") is to manually type in the broader taxon the autosuggests point towards. I would never consciously place something as unknown. I try not to use the autosuggest blindly for species level IDs full stop. But sometimes even I give in. I don’t think it’s wise to use species level autosuggests at all for insects… even in well covered locations like UK…it just tends to slow down the process for identifiers and make more work. This would be different again if the notifications were fixed so that we could see clearly when people disagreed with an ID…

@taylorse

A lot of work is going into developing the computer vision if its primary purpose is to support people who already have done significant research.

It may well reach a point at which species level IDs are reliable even for complex taxa.
It’s just not there yet. In the mean time, the issue is in the interface design, not the model itself or the contributions to developing it.

@taylorse

The person in your scenario does not need a list of suggestions. But for someone who is trying to learn, it’s a powerful tool.

Nobody knows all taxa at all locations. At home in the taxa I know well I am aware of the limits. When travelling or in taxa I know little about I am trying to learn like anyone else. So I regularly use the list of suggestions, but I use it to guide me to see what the options are once uploaded and then research whether a species level ID is plausible. Even if plausible I will take caution usually to place at a higher level unless it seems to be an absolute no-brainer.

Properly exploring options from the autosuggest during upload just isn’t a natural space to take the care needed for species level IDs in my opinion. But everyone is different. If you are only uploading a single observation or uploading birds,…then maybe it makes sense. If you are uploading 50 photos of 10 insects, then flipping into different browser tabs to do the research you would need to fully evaluate the autosuggest risks losing the content during the upload process.

On a phone, doing the research necessary to look into a species-level autosuggest is just completely impractical.

[taylorse]

I think we can agree that it would be helpful to make it easier to enter higher-level identifications without having to agree that we should hide less-confident suggestions.

In the app, at least on android, if you look at one of the suggestions you can select a stage higher up and make that the ID instead. It would be great to be able to do that on the website, too! I would also love it if it were easier to compare the taxonomies of all the suggestions, to see where they match. It is currently rather tedious to do this if you don’t already have a pretty good idea.

But all of this is for people who have a fair idea how to use the app already, and some idea how taxonomy works. It would be great if the computer suggestions could go back to a higher level, especially in cases where none of the suggestions is obviously the most confident. This would help people to learn about how taxonomy works while also making fewer mistakes.

[DianaStuder]

There is an open? request to display the taxonomy so we can choose the appropriate level.

[taylorse]

Oh I think I just found it! It appears to be “under review”, but I voted for it anyway because it sounds awesome.

https://forum.inaturalist.org/t/identify-page-observation-page-dropdown-list-of-upper-taxonomic-levels/12319

[jasonhernandez74]

@sbushes

If people are going to do said research to ID any observation, it is more likely to happen outside of the upload screen which is more of a transient space.

On the Android app, the autosuggestions are set up to facilitate comparing one’s own image with a set of images of a suggested taxon. I do that a lot. (And then I make sure to comment that I used the CV.)

@taylorse

If you have to do significant research, most people won’t identify anything. (And we don’t need iNaturalist for that. Most people were already not identifyng anything.)

I’ve said this before, but academic types saying “We’ll train them to think like us” is not a viable solution. There seems to be an undercurrent on the Forums of resistance to breaking down the academic/layperson barrier, at least when the subject is taxonomy or identification.

[upupa-epops]

@jasonhernandez74

I’ve said this before, but academic types saying “We’ll train them to think like us” is not a viable solution. There seems to be an undercurrent on the Forums of resistance to breaking down the academic/layperson barrier, at least when the subject is taxonomy or identification.

I think this tension/resistance is inevitable. You get the academic perspective on the forum, and the layperson perspective in the mass of observations that come in from laypeople.

How do you meaningfully connect the curiosity of a child with a phone and an academic researcher studying population dynamics, when they have substantially different and sometimes incompatible conceptions of basic terms like “tree”, “mushroom”, “fish”?

There’s going to need to be some flexibility on both ends. Is it worth creating a parallel taxonomy where a “tree” category exists? Or do you say that when you want to learn what kind of tree that is and how it’s related to other trees, the first step is learning that there is no such thing as a tree category?

[sbushes]

@taylorse

In the app, at least on android, if you look at one of the suggestions you can select a stage higher up and make that the ID instead

I’ve never used the Android app. It surprises me this feature exists there but not on the new iPhone app or in the browser. Would be great to have.

@jasonhernandez74

On the Android app, the autosuggestions are set up to facilitate comparing one’s own image with a set of images of a suggested taxon. I do that a lot. (And then I make sure to comment that I used the CV.)

Superficial comparison against images of suggested taxa doesn’t suffice with the vast majority of insects when going to species level. That’s the problem. When I talked of doing the research necessary, I meant beyond superficial comparison.

That said, incorrect species level IDs wouldn’t be such an issue if the broader algorithm functioned differently…

@jasonhernandez74

I’ve said this before, but academic types saying “We’ll train them to think like us” is not a viable solution. There seems to be an undercurrent on the Forums of resistance to breaking down the academic/layperson barrier, at least when the subject is taxonomy or identification.

I am a layperson who has had that barrier broken down through use of different platforms.
I don’t think the issues being discussed here are about things iNaturalist couldn’t change for the better. When I started recording, I used various platforms, each had it’s pros and cons in terms of breaking this barrier.

[DianaStuder]

First step would be - to somehow achieve better mentoring of willing newbies. If the problem ID was, long ago, from an observer or identifier who used iNat as a ‘flush and forget’ - not much to be done, but pile on the right IDs till the CID algorithm is convinced.
But if the newbie / or not so new is active and willing, then it is worth a brief comment for all future eyes on that obs.

I have worked thru the Geomodel Anomalies for the Cape Peninsula. I feel the bulk of those came from pre-Geo suggestions.
249 was doable. Next step would be Western Cape, but then I hit 10K, and I retreat in quiet despair. Will leave those to taxon specialists to pick thru.

[eyekosaeder]

@taylorse

The list of suggestions can also function as a list of similar species to rule out.

Yes, this is one of the most important CV features for me. If I do attempt a high-precision ID on one of my observations of a taxon I am not very familiar with, this is always my starting point.

@sbushes

If people don’t know they exist, they probably shouldn’t be adding a species level ID without significant research in any taxa…at the very least in more complex taxa other than birds.

I’m sorry if this is too epistemological… But if you didn’t know something existed, how would you know that you didn’t know it existed? It’s a paradox.
Even if you were confident that you know all species of a given taxa: Confidence and knowledge are separate scales. (I’ve probably been confidently incorrect an embarrassing amount of time)

Free/cheap information is hard to find if you aren’t affiliated with an institution. If you should only add a species level ID when a taxon expert (whether professional or hobbyist), then iNat would lose a lot of its usefulness to everyone. (Observers no longer get accurate IDs, researchers have to do more work if they want to use iNat data, etc.) So in my opinion the way it is now is a good compromise. If I have interpreted these results correctly, then RG observations are 95% accurate. I feel like any additional gain in accuracy would come at the cost of having a lot fewer IDs.

[DianaStuder]

If, you require, every ID to be - I stake my professional reputation as a scientist focused on …
Consider where we would be with IDs for the Great Southern Bioblitz - and that is with the commitment and enthusiasm from a bioblitz.

263K obs from 8.6K observers
with 4.5K identifiers.
We have 48% RG, a negligible 3% broken or Not Wild.
Observations distributed across 52% green stuff, 24% ‘bugs’ and the last quarter scattered across all the ‘iconic taxa’ that ‘people recognise’.
130K obs still at Needs ID for dedicated scientists.

If you would restrict those IDs to what scientists can squeeze in around a day job, iNat will dwindle to - oh yes, I remember when that was buzzing with activity! Wonder what happened to it?

[miriameaglemon]

@taylorse

The alternative is to leave everything you don’t know about in “Unknown”, which is a great way to make sure that nobody who knows more than you ever sees it. Putting in an incorrect ID that is at least in the right ballpark gets it in front of people who can help you.

I’m glad to see this; personally, I’ll give my “best guess” ID on a tricky moth if I’m not 100% confident on the ID, and I WANT someone to “disagree” that it’s a so-and-so if it’s not really a so-and-so! My feelings aren’t hurt at all - that’s how I learn!

[upupa-epops]

This gets controversial (depending on how much effort is put into the “best guess”):

@schizoform

Was it an intentionally wrong identification, or an intentionally reckless guess? I’ve done the latter (with awareness of Cunningham’s Law!) and don’t think it’s necessarily bad practice. Pointing out that an intentionally wrong ID can be misleading should get them to stop.

[taylorse]

@miriameaglemon

I’m glad to see this; personally, I’ll give my “best guess” ID on a tricky moth if I’m not 100% confident on the ID, and I WANT someone to “disagree” that it’s a so-and-so if it’s not really a so-and-so! My feelings aren’t hurt at all - that’s how I learn!

To me the basis for this is that people who are learning will make mistakes. (People who are experts will also make mistakes!)

I don’t think we should exactly encourage people to make mistakes, and certainly we don’t want people to just make wild, thoughtless guesses, but fear of of making mistakes is a major obstacle to learning, especially for adults, who often don’t have to learn.

Therefore it needs to be okay to make mistakes, or many people won’t even try to learn. And as you learn, you make fewer mistakes (or at least different ones).