Why You Should Add IDs to Research Grade Observations

Hello, I just wanted to make a quick post encouraging people who are confidant about their identification skills to also, when they have time, go through observations that are already Research Grade and confirm their IDs too, not just ones that Need ID!

It’s very simple: Sometimes observations that are Research Grade are very, very wrong, usually because the original identification added was made with the computer vision, (mostly the version from 4+ years ago) which suggested an incorrect but similar looking species, then one other person came and agreed with that incorrect ID.

Some of them are very obviously wrong, but others, the details that point to the correct ID take careful examination of the picture.

I’ve got a self-appointed goal of adding Plant Phenology annotations to all observations of pawpaws, so starting with the oldest observations first, I’ve been going through observations of pawpaws adding annotations and confirming or correcting the ID when I’m able to.

A lot of the mistaken identifications are from the older computer vision models, but every now and then it’s just a normal case of people just mixing up the species.

So I would like to take this opportunity to encourage more people to include Research Grade observations in their identification sprees, just to make sure they are actually correct :) And if it is, three agreeing IDs is better than just two!

Anyways, check out the plant phenology graph for pawpaws and see all the progress I’ve made :)

31 Likes

Hear, hear!

I also regularly go through RG critters to check them out. I find some misidentifications but the vast majority are IDs which lack sufficient evidence. So it’s a good opportunity to fix them up.

The good thing about iNat is that no observation status is set in stone. Moving them from RG to NI isn’t a backward step, it’s about making the data as good and trustworthy as it can possibly be.

There are of course many other good reasons to add IDs to RG submissions, even if they are supportive. In another recent thread there has been some debate about the value of doing this but I don’t agree with the naysayers. A considered ID is of equal value whether it’s the first, fifth or one hundredth. So I would encourage people to add their considered IDs (please don’t blindly guess, and don’t trust the machine learning!) no matter how many already exist.

13 Likes

Nice! Congrats! It´s lovely to see a pattern emerge :-)

I do go through certain RG observations from time to time, but I feel as long as there are so many non RG observations waiting to get at least one additional ID or even get moved to RG, that my time is spend ther more valuebaly.
However, when I do a deep dive in a species, I often end up looking at the RG observations at one point as well. I will usually only quicly skim through the tiny pictures and pic the ones out, that look weird to have a closer look and correct if needed. I will also go through the outliers on the map or the ones that are at the rim of the range to check them more thoroughly. Almost every time I decide to do that I will find some that need correction, so it is for sure worth ones time and it kind of feels extra rewarding.

10 Likes

Thanks for doing the phenology annotations! I got tired of doing them with IDs but you motivate me to start again. It’s very valuable.

4 Likes

An important reason to add an ID to a Research Grade observation: If an iNatter deletes his/her account, all the ID’s that person did go away! Lots of things that were RG become Needs ID.

So I would say, don’t search out RG observations just to do this (too many Needs ID need your help), but if you visit an RG observation for any reason and you can ID it, add your agreement!

17 Likes

I would also try looking at RG genus-level observations. Often, they can become RG at the genus since a species taxon wasn’t available on iNat at the time, especially for older observations. You can go in and improve the community ID to an observation which would otherwise be stuck at the genus.

7 Likes

Neat work on the paw-paws. I hadn’t realized the phenology graph could be so detailed. You’ve inspired me to be more active about adding those details when I ID.

5 Likes

Don’t certain external datasets/researchers also use higher agreement thresholds for determining what is/isn’t “research grade” (for instance a researcher could choose that they only want to use observations that have three agreeing IDs for their dataset because this would lead to even higher accuracy than the two that iNat uses, albeit at the cost of a smaller set), which means that adding ‘redundant’ IDs to things that are already RG could still actually have some use?

1 Like

I am not sure if any researcher would just rely on those community IDs? I for sure would not and would check each of those observations again, whether they are indeed the species I am interested in…

7 Likes

I thought this 3rd ID was silly, and I was wrong. Thanks for posting.

8 Likes

The majority of my IDs result from going through observations of taxa that catch my interest for one reason or another. When I do this, I usually don’t distinguish between “Needs ID” and “Research Grade” observations. Both have equal potential to be relevant to my interests.

However, I usually don’t add IDs if I believe the current ID is correct and it’s already at “Research Grade”. I find it a bit annoying when I encounter an incorrectly IDed observation that has a pile of incorrect IDs on it, so that when I add a correcting ID the observation just stays at “Research Grade” under the incorrect ID. It doesn’t happen often and it’s not a big deal, but it’s a minor irritation. I prefer to reduce the likelihood that I am creating that experience for someone else.

6 Likes

Of course one should still be rather confident of the Id and not just agree because several others did.

…In the case that it stays at RG… use the “can be improved” helps…

2 Likes

There are two kinds of people: those who know they make mistakes, and those who don’t know that they make mistakes. :-)

4 Likes

If I didn’t know I make mistakes, my notifications would make that clear!

Nonetheless, I strongly recommend adding a third (or other) ID to RG observations because of the negative effects if an identifier deletes his/her account – or has it deleted by a well-meaning relative after death.

6 Likes

Yes, many researchers do use Community IDs for RG observations as these are what gets exported to GBIF.

2 Likes

Lots of observers are using garbage-bin-species; usually species that are well-known, and are well presented with photo recognition… In the fungal world such species are e.g. Phellinus igniarius, Amanita muscaria and Flammulina velutipes. These include lots of wrong RG-level observations, and it is good to check them.

One problem are the followers: when someone gives an ID to their observation, they agree fast, without knowing anything of the species. It is not a big problem if the first ID comes from an expert, but when it does not, we get wrong RGs. So, we should be very well aware what we do know, and what we don’t, and act accordingly.

2 Likes

You used to be able to add such taxa to the list of problematic species here: https://forum.inaturalist.org/t/computer-vision-clean-up-archive/7281 But it appears the wiki is no longer editable.

Thank you for the work with Plant Phenology! Also, I agree with many of your points, and the ones in the comments. The one about deleted accounts makes a ton of sense. I’ll be linking people to this post when they get mad at me for a confirming ID, haha.

3 Likes

That could just mean that other people make mistakes. :-)

Once in a while they’re wrong, but I’m followed by some pretty good identifiers and usually the mistake is mine.

3 Likes