I think that’s the key. If 5 knowledgeable people all agree on an ID, then it’s very likely to be correct. I’m sure we can all find some isolated examples where all the experts were tricked, but it’s uncommon.
IDs lacking critical thought are bad though, so of course 5 of those is extra-bad.
I think the conclusion really should be as simple as “extra good IDs are good, but extra bad IDs are bad.”
I studied this issue in my research species. While I didn’t worry about the “knowledgeable” or not, I did find agreement with what you say about the number of IDs. Here’s the relevant figure from that paper showing that community identification (CID) reaches near 100% accurate at four supporting IDs (five IDs in total).
The reference for the paper is:
Zani, P. A. 2024. Accuracy of iNaturalist identifications: a test using a geographically ubiquitous and variable species of lizard. Herpetological Review 55:157-162.
This. The original post is not asking about identifiers adding additional IDs generally to observations within their area of expertise (which is useful), it’s asking about the phenomenon of people adding large numbers of superfluous IDs to interesting observations, such as observation of the week, often as a form of “like” (which is, at best, irritating). There are likely more confirming IDs on that moth observation than people in the world who know how to ID the species, and the same is true for extinct curlews, tree frogs, and so on. I’ve seen Tony push back gently against this occasionally on observations of the week, but without much success.
Removing the “agree” button on observations that have more than a certain % of agreeing IDs (90%?) would likely do a lot to help.
Now what? Tag other people from the leaderboard or who you know ID in your area. Message the other wrong identifiers, explain why you think the ID should be changed, and ask them to change. (Include the link to the observation!) Be patient.
Removing the agree button was tried briefly – briefly because it was loudly shouted down. Of course you’re suggesting it only for observations with several agreeing ID’s already, but I doubt this would ever fly.
Except that there isn’t an easy way to tell an extra “good” ID from an extra “bad” one – “bad” IDs are generally made in good faith, even with conviction.
Sometimes our collective knowledge may change – people have been assuming that only species X is present in a region and IDing observations as such, until someone realizes that there is a second similar species Y and that many of the existing observations need to be reviewed and some portion of them changed. Or there is one case I can think of where many people were relying on one particular influential (and otherwise good) resource that happened to have a mistake such that a particular set of characteristics keys out to species A even though it should be species B. In other words, they were applying the wrong name to the right species concept.
The more IDs an observation has, the more difficult it is to fix such issues when they occur. Yes, we can tag people and call in more users to help, but in taxa where there is a scarcity of IDers in the first place, this is not a particularly good use of time and resources.
(I mean, do what you want. I recognize that some people have their own reasons for reviewing RG observations. But adding additional agreeing IDs merely as insurance against an IDer deleting their account seems like the wrong solution to that particular problem. What is needed is to reduce the number of account deletions (and ID deletions) that happen in the first place.)
You shouldn’t agree on observations that “seem correct”! You should ID observations that you are confident identifying and understand why it is this and not something else.
Its an interesting graph. But if 997 out of 1000 initial identifications were correct, that suggests it is an easy lizard to identify, even though it is variable. Has similar analysis been done on more difficult species? I suppose you then get into the question of who can say for certain what the correct identification is.
Even where the identification is 100% certain, wouldn’t the scarcity of identifiers (brave enough to suggest a meaningful ID) prove an issue?
Yes. Maybe one could choose species that are difficult but not impossible from external characters but have clearly different genitalia, such as many moths and beetles. You identify the specimens from the genitalia but don’t include them in the uploaded photos. There would probably be enough people willing to give an opinion on the moths but you might struggle to get enough coleopterists.
How do you propose to do that? As has been discussed elsewhere, it seems to be iNat policy that a user has the right to delete their account along with all their IDs and comments. If that ability isn’t changed in some way (e.g., by keeping the IDs and comments but removing the inatter’s username), then a few extra confirming IDs on a RG record seems the best approach. I tend to add an extra confirming ID on RG records with only 2-3 agreeing IDs for that reason.
There have been various ideas suggested in previous discussions: for example, providing users with an option to deactivate or lock instead of delete their account; implementing a “cool-down” period before account deletions are final; better mediation in cases where account deletion is the result of ongoing conflict; some kind of pseudonymization/anonymization if an option can be found that would allow this without compromising the integrity of the IDs associated with the account, etc.
I strongly believe – as I have said before – that account deletion is a problem that urgently needs to be addressed better than it is at present.
Adding extra IDs to an observation doesn’t resolve the underlying issues and disruption connected with account deletion – there is a lot more that is lost than just IDs. All it does is (somewhat) soften the impact. Again, I’m not saying don’t do it. But it should not be confused with a solution and it is not a replacement for a proper solution.
I realize that IDs are something we can control and changes to how iNat works are up to staff and developers – but it seems like we are doing ourselves a disservice if we focus on compensating for the problem by adding more IDs instead of insisting that we need a better fix.
Given that there is no way to predict who might delete their account at some point in the future and what taxa/observations will be affected, I don’t personally feel that adding extra IDs merely as insurance against an eventuality that might or might not come to pass is a particularly good use of my time when there are thousands of observations that have not been reviewed by anyone. If at some point there is an account deletion, the observations will become “needs ID” again and will simply be added back to the existing pile. The calculation may be different if you happen to work on a taxon where the shortage of IDing capacity is not quite so acute.
I don’t disagree with what you’re saying. If someone who does bird IDs deletes their account, it will likely have little impact on the bird records they added IDs to since there are so many bird IDers. Probably a much bigger issue for many invertebrates and plants.
I strongly think we should add agreeing ID’s to RG observations if we’re looking at the observation for some other reason. Searching out observations just to add more ID’s seems a waste, but we look at already-ID’d observations for many different reasons. Let’s add an ID while we’re there, if we are confident we know what it is.
[I failed to express what I meant. Another attempt below.]
Well, one person’s waste of time on iNat might be another person’s way of enjoying and learning from the site. No one is obligated to review records in a certain “ideal” way as envisioned by someone else.
Too little potential benefit (“what if someone deletes?”), too many potential drawbacks (endless notifications + ending in leaderboards). Your Mileage May Vary.
If you want to reduce the agreeing wrong IDs, make sure you write a comment saying exactly why. And the rest of us learn and will probably check a few of our own similar observations.
The part in parentheses is what is proving an issue. The general trend of this and previous threads on this topic is, “don’t be brave enough to suggest a meaningful ID.”
Except that sometimes it can take years before people come along and find said wrong IDs. And by that time, identifiers may have become inactive - which makes wrong IDs hard to fix, particularly if there are too many of them.
(post deleted by author)
