Computer vision IDs should not be eligible for Community ID opt-outs

Everyone on this thread makes a great point. However, I am adding another layer of icing on the cake to clarify further. Some people on Inaturalist use Inaturalist for research, some for projects for college, some for fun, and a select few for foolishness. Professionals and experts still use AI on certain organisms. This is not a problem at all. Beginner naturalists use AI sometimes as a quick getaway. Remember, everyone was a beginner once, and to be honest, I was a pretty bad beginner. I…
made a secondary account to get my observations to RG.
used crazy AI options for observing AND IDING.
speed identified taxon that I did not know about well enough
stringed organisms(one occassion calling a SESA in breeding colors a REPH.)
asking photos that were ok to be Observation of the Day.
I have not done most of those things in months. I STILL use AI but type the words in instead. And also overconfidence in ID’s. I still am learning, and everyone is, whether it be the beginners like me in May, an amateur like me now, or the world’s best expert on botany.
Robby

4 Likes

For this reason, I do think AI users should be eligible for community ID opt-outs

1 Like

Hey @bazwal. iNat is a pretty friendly community. I think we can kick around ideas here without taking this personally.

My experience of IDing a whole lot of observations is that for every user who has chosen to opt-out of Community ID with clear knowledge of the implications, there are many more who didn’t realize the effect of enabling this option. Realistically, a user would need to have accumulated a fair bit of time using iNat in order to understand the concept of Community ID, and yet quite a lot of the “opt-outs” are from people who have very few observations. Many of those users haven’t been active in a while.

Someone making an informed choice to opt-out of Community ID is essentially saying “I’m confident in my identification of this organism (or all my IDs) and I don’t want that changed based on the consensus of other iNat users.” That’s fine, but it’s logically not compatible with “I didn’t know what this was so I used computer vision to tell me.”

As some have pointed out, the way the UI works means that it’s possible that a knowledgeable user who’s opposed to Community ID may have used computer vision to select the ID that they don’t want changed. To me this seems like something of an edge case. It’s also clear that, if iNat were to implement logic that suppressed the opt-out for computer vision IDs, this hypothetical observer has any easy option of editing the ID, pressing a key, pressing backspace and clicking Save.

For every opt-out aficionado who might occasionally experience mild inconvenience, there will be many novice users whose observations get accurately IDed more quickly and don’t have to scratch their heads wondering why the suggested IDs don’t change the overall ID listed for their observations. Also, among orphaned observations, there will also be a significantly lower proportion that have incorrect IDs that will never be fixed. That’s a secondary aim of iNat, but still a valuable one.

To my mind, someone who wants to opt-out of community ID should be able to do that when they affirmatively provide their own ID. But setting up iNat to prioritize computer vision IDs over community knowledge isn’t helping anyone.

8 Likes

It’s best to avoid attributing motivations to people based on your perceptions of their tone. I actually have no personal stake in this at all, because I very rarely use the Computer Vision and have never opted out of a Community ID. My only concern here is with how the proposed change affects other users. After the debacle of the Agree Button removal, we all need to be more aware of that.

4 Likes

I agree. If I read your comments wrongly, I apologize.

4 Likes

The problem here is that a lot of observations don’t get that top suggestion. The CV seems to have gotten a lot stupider lately, so I’ll show it a chickadee or junco and it’ll come up with 10 species-level, completely wrong IDs, and nothing else. If I honestly didn’t know what the bird was, I’d have a problem.


Okay, never mind. I think it’s had an update since the last time I submitted observations, because it now gives me similar silly suggestions, but actually tells me it’s not very confident. Hmm.

2 Likes

I think a middle-ground would be that observations from opted-out users be made casual if they disagree with the Community ID, perhaps at a 3-1 standard. Bottom line is that there’s no reason for opted-out, evidently incorrect IDs to sit at ‘Needs ID’ forever. There are many reasons to keep the option for individual observations, though.

In personal experience I see most opted-out users are either A) dead accounts with 3 obs. (90% of cases) B) 10k+ obs. accounts but never reply to pings or suggestions. Only in rare cases do a see an opted-out wrong ID corrected by the user.

As an example, I recently corrected all the observations of Galápagos Finches reported on the wrong islands (because of computer vision). One user had reported many finches on islands they don’t occur on, but had opted-out of community taxon. When I corrected them, I clicked the “Community Taxon - No, it’s as good as it can be” box to make them casual and remove from range maps. The user then undid that to make it “Needs ID” again and blocked me.

There’s no reason for observations like these to sit at ‘Needs ID’ and clog up the Identify section for users trying to identify things for people. This just wastes their time and hides legitimate observations in the slough of opted-out and blurry distant photos that make up the heavy majority of ‘Needs ID’ observations of certain taxa.

An expansion of the use of the “Based on the evidence, can the Community Taxon still be confirmed or improved?” box to encourage the pruning of bad-quality observations would result in the ‘Needs ID’ observations being the recent observations that need identifying – Not blurry photos of birds taken with an iPhone 200 ft away and opted-out obs.

EDIT: Blocked out username

10 Likes

@cgbc just as a reminder, the forum rules don’t allow bringing issues about specific iNat users into the forum, so i think it would be better if you deleted that photo or blurred the name there. If you think someone blocked you to avoid your ID on their observation, which is against the iNat guidelines, you should email help@inaturalist.org .

4 Likes

Only loosely related to this topic: Should it be allowed to opt out with observations without any ID at all? These may stay as ‘unknown’ forever. There might be legitimate use cases, no idea, but it does annoy me when IDing unknowns I have to admit.

5 Likes

Actually, I had to opt out a while back. I had someone put an absolutely ludicrous ID on a Bombus affinis observation, sending it to Genus limbo. So I opted out for a couple of months until I got a couple more ID’s to get it back to species. Currently, it’s opted in.

3 Likes

I see why you clicked the ‘Community taxon can’t be improved’ button to clear up the range maps, but I’m slightly conflicted, because of course you do believe the community taxon can be improved - in your example it is at genus, but you clearly think it can be identified to species. The problem, as you say, is that the observation taxon won’t follow it, and so it stays on the range maps and clogs up the NeedsID. So what else to do?

I liked a suggestion that I think I read, but now can’t find, that opted-out observations should become casual at the point they would otherwise have become research grade with a community ID that disagrees with the observation taxon. I think that would deal with the underlying concerns of this thread, and make it unnecessary to force things into casual in an artificial way.

1 Like

Legislation does not cure problems, but education can improve the outcome.

I believe once the disagreeing poster’s ID becomes Maverick the observation becomes casual, right?

4 Likes

For a brief moment I misread the title of this request as " All IDs should not be eligible for Community ID Opt-outs" and thought it referred to not allowing opt-outs for, say Order or Class level IDs (e.g. someone adds a “bird” but has a blanket opt-out on all of their observations, so ID can never be confirmed). Is there a request anywhere like this? Certainly, it should not be possible to opt-out if you have not even added an ID, and yet this is possible, meaning that such observations can never be confirmed.

Question: Does anyone really need to have opt-outs at a level above, say, genus? If such a feature request doesn’t exist perhaps we can split this off and can add one.

The preferred terminology is “computer vision” so I updated the topic title.

1 Like

Right you are. I thought I’d seen observations with a lot more contradicting IDs than that still in the NeedsID pile, but I must be making it up, or there must have been some other reason. In that case I don’t really understand what the issue is. If people want to opt out they may, and if the community disagrees with them their observation becomes casual and disappears. The ID is no more difficult to ‘correct’ than any other, is it?

But it makes more ID than usual to for opted-out IDs to become maverick.

1 Like

Interesting, thanks. This has clearly passed me by. What is the ratio then, and what is the rationale?

I assume @raymie just means that an opted-out observations with an incorrect ID requires 3 identifiers to agree on a different ID before the observer’s ID is regarded as maverick, whereas only 1 identifier is needed to agree with the observer for a correct ID to reach RG. That’s just a consequence of iNat’s general rule that community IDs require at least 2 IDs of which more than two-thirds must agree.

1 Like

I’m sorry, I misspoke. Yes, it only takes 3 IDs to make a maverick on an op-out observation. However it takes more than that (I can’t remember the exact number) for the observation to become casual, otherwise it just sits at Needs ID.

1 Like