Opt out of AI identification

We all know that careless use of the CV by observers can cause problems and I fear the only solution is probably better communication, maybe in the form of popups at least for newish users. I’m wondering though how much the specific behaviour raised by @davidhowdon represents a serious issue in numerical terms. My impression is that the vast majority of IDers likely to add a second ID to an observer’s original ID are already regular iNat users who are probably well aware of the difficulties a hasty ID can create and open to communication with the observer. Perhaps a simple comment on the observation indicating the observer’s doubts/wishes would be enough to make the subsequent IDer pause and think before automatically hitting “agree” or clicking on the first suggestion. And if that doesn’t work, direct communication with the “guilty” IDer asking them to explain/reconsider/withdraw should be enough. I know that when I’m in doubt, I may “suggest” an ID, but add “possibly” or simply “???” in the comments to make it clear that it is only a suggestion. I am though 100% sure that, as an IDer, if I weren’t able to use the CV suggestions as a quick way to enter a name, I’d probably give up IDing altogether. Life’s just too short to turn IDing into a slow and taxing spelling challenge.

Mostly, yes, I also find that the majority of people who make IDs for others do so responsibly (occasionally this does include making IDs based on certain common misconceptions, such as what label should be applied to dandelions, but in general most people who suggest IDs seem to be wise enough to recognize what they know and don’t know).

But it does seem to happen semi-regularly (I see it a maybe few times a year in the area where I am active) that an enthusiastic newish user will start going through observations and providing IDs based on a combination of the CV suggestions, image comparison, and half-knowledge. Usually what happens is fairly quickly they start getting pushback and people disagreeing with them and encouraging them to change their IDing behavior; in most cases this has the desired effect before too much damage has been done.

I find this is always a bit tricky to navigate – on the one hand, we badly need IDers; on the other, someone who is guessing IDs can have a fairly substantial negative impact in a short amount of time and it is often quite time-consuming for the taxon specialists to subsequently correct the damage. It is not always easy to figure out a strategy to encourage such users to channel their enthusiasm in a more knowledgeable manner while not discouraging them from IDing. I fear I do not always react well to people putting IDs on my observations when they clearly have no knowledge about the taxon in question, and this has probably led to me making a few people feel unwelcome who might have become valuable IDers if a more gentle approach had been taken.

I feel like I keep saying this, but I think a key part of addressing users’ tendency to irresponsibly apply the CV suggestions (whether on their own observations or other people’s) is teaching them how to use it responsibly and to critically evaluate IDs. I’d like to see this instruction embedded in the interface, not merely in the form of optional tutorials that most people will not find or watch. I realize that this is a huge ask and a substantial challenge to not only develop an appropriate pedagogical strategy and materials but figure out how to integrate it into the interface, but I think it is the only effective and sustainable way to address the underlying dynamic. Tweaking the CV suggestion mechanism or making it more difficult to access the CV will only treat the symptoms, not the causes.

We have a project – built by a local entomologist
https://www.inaturalist.org/projects/not-in-sa-insect-cv-clean-up
I use that to clear obs in the Western Cape.

https://www.inaturalist.org/posts/134403-help-clean-up-south-africa-s-insect-data-new-project-not-in-sa-insect-cv-clean-up

This random obs from the project is in Kwazulu-Natal (and out of my range)
https://www.inaturalist.org/observations/401806736

For my local obs, I try for a better / local ID, leave a comment and @mention as appropriate.
Keep trying https://www.inaturalist.org/observations/363595361
Success ! We are now at a local Genus, and over to taxon specialists for sp

NOT calling out ‘bad’ behaviour. (But the project link is the important one) Showing how I explain why the ID is wrong. I used to ask – please check distribution maps – but that prompt seemed dead in the water.

If the iNatter is new, or active – it is worth nudging them in the right direction, while they are open to new info.

I think you hit the nail on the head. It is unclear to me who’s job is to provide solutions.

That was me.
In my defence, iNat did not direct me to the keys, some patient members of the community did.

Glad you stuck around, accepted correction, and were willing to put in the work to learn and make better IDs!

Feel free to ignore me if you wish. I’m one of the people who starts typing a name and then clicks on the correct item from the list because it’s easier to type “Goniofusus d” than remember how to spell “dupetitthouarsi.”

(if you start typing then select from the list of suggestions your ID will not have the CV badge. as soon as you start typing the suggestions are not based off the CV at all)

The other thing that I was trying to get at in my previous post is that people tend to find being corrected rather uncomfortable (particularly if they are aware at some level that they were not being as careful as they should), no matter how kindly and patiently the correction is presented. So it is quite challenging for IDers to find an approach that will not make the person become defensive or feel that they have lost face.

One of the reasons I would like to see an embedded strategy for helping people think about their IDs is that it would make the process of being guided and corrected more neutral – users would be less likely to feel that they are being personally criticized if the feedback is coming from the website and not from a particular person.

That is true. Though learning to realise you are sometimes wrong and being able to deal with that is a bit of a life skill isn’t it. If I think someone’s put the wrong name, I try to explain the features that make it what I think it is, and the features that mean it isn’t what they think it is. I try to do that in a polite and supportive way. (And I may still be wrong of course). That’d be the approach you’d take in real life one hopes. But if the recipient still takes that help as personal criticism I do believe that’s a problem for them to deal with inside their heads, not for us IDers or the website. I only comment because of a recent experience where someone was a bit defensive that their ID was correct, despite various people’s polite pointers to why the species was something else. If those pointers were coming from ‘the website’ I still think that observer would have dismissed them.

Indeed, even to the point where using the CV, clicking on the CV-generated name that appeared in the search bar, and re-selecting the same species when it appears in the drop-down menu will make the CV badge not appear on your ID. I do this sometimes when I’m sure the CV is correct but I’m IDing for someone who hates to see the CV badge. The presence/absence of the CV icon truly says nothing about the identifier’s actual knowledge (or lack thereof) about the ID they’ve added. Sure, some newer users click the CV mindlessly, but plenty of users copy wildly incorrect IDs from Google Lens, or type in outdated names from old print literature, or use any other number of poor ID sources that don’t cause the CV badge to appear.

Getting it wrong is a good life lesson and good training. We all get it wrong at least occasionally. Although it’s uncomfortable, it can make us better at making future decisions.

I mean, yes, I agree, but the unhappiness of the person being corrected may end up directed at the person who is doing the correcting, which most of us don’t find enjoyable even if we know it is probably not anger with us personally.

In addition, not everyone is skilled at being diplomatic when providing corrections, particularly if it is a case where one needs to not only correct a single ID but encourage different behavior following a series of unknowledgeable IDs. Or one’s ability to be polite and supportive may be lower than usual on a given day. Or there may be intercultural differences in the expectations about the way criticism should be communicated. We may feel a certain responsibility that if only we had communicated differently the user would have responded better.

Of course, navigating such interpersonal matters is no doubt also a useful life skill to develop – but it is also emotional labor that could, in some percentage of cases, likely be avoided by having better integrated onboarding mechanisms.

In other words: this is at least as much about sparing the corrector as sparing the correctee!

In essence, it’s a teacher-student relationship, with all the complexity you find in that in a classroom, but being conducted informally and remotely. It’s never going to be simple.

Well, except that the students may not see themselves as students…

(And I suspect some identifiers do not see themselves as having volunteered to teach…)

It’s a helpful way of framing it, though, and perhaps it would make sense for iNat to more explicitly communicate its mission in terms of shared learning and exchange of knowledge (which is not identical to being a platform sometimes used by educators) – I think some of the interactions that I have seen go badly wrong are at least partially the result of one or more parties viewing their role according to some other paradigm.

It can also mean that they know the species but using the CV to add it is faster.

Returning back a bit to the original topic of the thread, I could see some potential advantages to a somewhat more sparing and user-controlled integration of the CV suggestions into iNat – albeit not in the form suggested in the initial post.

When I saw the thread title, I thought it was going to be about wanting to opt out of automatic suggestions when entering IDs oneself. At present, there is no way to turn off suggestions in most interfaces (Identify being a notable exception) – you can override them by typing something yourself, but you can’t prevent the system from trying to provide them in fields where this function is active. Occasionally in the forum people have noted that they would prefer to come up with IDs themselves, whether for their own learning processes or because of a dislike of AI, and therefore they find the suggestions unwanted and intrusive.

So a user-level opt-out option of “don’t show me CV suggestions” or “only show me CV suggestions if I request it” might be a way to give people more control.

I could also potentially see an argument for adding some minor restrictions on accessing the CV for identifying other people’s observations for users who have few or no previous IDs or who have a track record of a lot of subsequent disagreements with their IDs. That is, to encourage people to view IDing as something that should be based primarily on one’s own knowledge, they would not get CV suggestions on observations other than their own until they have made a certain number of leading/improving IDs for other people. I could see this as being a bit tricky to implement and potentially a source of confusion (why do I get suggestions sometimes and not others) and I don’t know if it would in fact discourage people from adding unknowledgeable IDs (maybe they would just use Google Lens or some other CV tool to suggest IDs) but I am putting the idea out there in the interest of brainstorming.

That would be me.

I don’t have extensive knowledge in my observed taxons, but I do have a field guide as a reference. I guess that counts as half-knowledge because that’s literally the extent of my knowledge as an amateur. CV helps tremendously for further checking, but I know my reference can be wrong or outdated so corrections are always welcome. But I found the more frequently happening case is that some people are (possibly?) rubber stamping Agree on my IDs.

That being said, I do agree that the UI/UX can be improved and the CV suggestions might exacerbate the issue for new users. When I was a brand new user, I was not aware that you can put coarse ID so I marked everything down to species level because that’s what the CV is suggesting. I suppose that misunderstanding (that putting a coarse ID is not considered a faux pas) is also one reason why this FAQ exists.

It actually is not the situation I was referring to at all – it absolutely isn’t necessary for everyone who provides IDs to be an expert and I wasn’t trying to suggest that amateurs can’t be knowledgeable. What matters is having good judgement about when not to select a highly specific ID, an awareness of the limits of one’s own knowledge, and willingness to do research to get more information or check when something doesn’t make sense.

I was referring to people who are relying primarily on the CV and what “looks right” when IDing other people’s observations; often these people suggest species-level IDs for a wide variety of difficult taxa and when asked, it generally turns out that they have not consulted any resources at all and they cannot explain why they chose the particular ID or how they excluded similar species. In many cases it seems to be motivated by interest and enthusiasm for all the interesting organisms out there, and suggesting IDs that they don’t have sufficient skill to responsibly provide is a somewhat misguided way of engaging.

Of course, don’t worry, I get what you meant. I just wanted to put some perspective from a somewhat new user (which hopefully helps the team improve their user onboarding workflow).