Why You Should Add IDs to Research Grade Observations

The idea is that a researcher is NOT going to RELY on community IDs, but they MIGHT form a starting dataset to look more closely at, by starting with observations that have more than TWO people confirming. There are taxa that lots of bad IDs, to find something that they MIGHT be able to use in their own study, start with three or more could lead to a better population to work through for further study.

3 Likes

As an added benefit, taking incorrect identifications out of research-grade helps the computer vision work better, leading to fewer errors in the future!

This is important too - a very prolific identifier deleted their account a while back and all of the things they took to research grade went back to “needs id”. A few more IDs on some of those could have been useful!

But of course, please keep on top of your notifications if you do this. There’s nothing more frustrating than trying to correct an ID that 5 other people have already incorrectly confirmed.

3 Likes

It depends on the research. If the researcher is working on one species or a small number of them, it may be viable and necessary to check the IDs. But if they are using millions of observations of many taxa to look at broader biodiversity patterns, that would be impossible, and in such a case the errors are likely to be harmless noise. Every dataset has an error rate, so all we can do is to try and keep it as low as possible for the observations we have control over or interest in.

5 Likes

Absolutely agree. Be they intentional or not, a small but notable number of false observations do make it to RG regularly often. I’ve been trying to check out RG observations as much as Needs ID ones a bit recently, and I really value the users here who’ve been doing that for much longer than I have.

By the way, when I’ve checked both, the RG observations have a higher percentage of correct ID than the “Needs ID” records we spend so much time on, often a much higher percent correct.

4 Likes

I’m currently looking over Research Grade Ptilidium observations and found a few incorrect ones. The observer was thankful that I corrected them.

7 Likes

As a specialist in the plant family Costaceae I filter for that family and regularly review any observation in that family and I agree it is important to add your “agree” even if it is already in Research Grade. By far the biggest problem I have seen is that the system defaults to “wild” and I see many, many RG observations that are clearly in cultivation - some in botanical gardens, some in climates where these tropical plants could not survive outdoors, and of course many that are far outside their native range and would only be “wild” if they had escaped cultivation. I even have had some “wild” that are growing along the street or in containers that you could see in the photograph. I REALLY WISH INATURALIST WOULD CHANGE THE DEFAULT TO NOT WILD! Too many people ignore the check marks along the bottom.

Also just to note - when you’re adding your IDs, check out the obs annotations and make sure they’re as good as they can be. Lots of people (myself included) don’t always bother setting these. I think it’s probably because they aren’t always in plain sight, but they can be very useful especially as searching filters.

1 Like

That would send the wrong message. It would imply that most observations are not wild, which is the opposite of what we want to encourage. It makes the most sense for the expected norm to be the default and the exceptions to have to be noted as such.

7 Likes

I wish there was a way for people to like an ID rather than agree. I know I wanted to hit agree when I was new because of training in other social media where <3 is the norm. It is a hard habit to break. Also upping the points somehow from people that do know what they are doing would help. Would people be able to self describe their level of accuracy or could that become part of the program? After X number of correct IDs, you earn a star or badge that lets your IDs weigh more in the percentages?

I write to people about this all the time. In USA bees the eastern species are well known and somehow are ending up clear out in Washington state trying to name our species and aree plain wrong. Usually someone else does see my plea and they start to help out which is a very nice result. I’ve met great people this way too.

2 Likes

Has come up often before in the Forum.
iNat treats all IDs as equal, and staff will not change that.

We have to learn to evaluate an identifier and their ID. What they choose to observe, or to ID, what they write on their profile, the snarky or helpful comments they leave.

2 Likes

The only way to determine “correct” IDs is the emerging community ID. Which is basically a popularity contest.

I think it is good to encourage people to add more confirming IDs, but in a sense it does make disagreeing IDs more difficult. On observations that have 3 or more agreeing IDs that I know are incorrect, My maverick ID doesn’t seem to do much to encourage anyone to check their IDs since it doesn’t change the status of the observation, and I seem to accumulate Maverick IDs because of this.

1 Like

As the original post states, this is a practice only for people who are confident to ID. Garbage IDs such as those from people looking for a proxy Like button are not helpful. However IDs made with consideration are, which is what we are talking about here. In theory, considered IDs should be moving towards the correct taxon, therefore the incidence of valid mavericks should reduce. If the IDs aren’t doing that, then they aren’t sufficiently considered.

2 Likes

That makes you one of the proud mavericks. Tag in someone competent to help move the ID at least to a broader taxon, where more identifiers will see it?

2 Likes

You need to click “can be improved”.

5 Likes

If I run into these, I will tag some other IDers I trust to help me out. It can also make sense to tag those people that provided the wrong ID, if they are still avtive, and usually at least part of them will retract their former ID, especially if my new suggestion come with an explanation. And for rhe time it is not yet resolved, I use the “can be improved” like Marina suggested. I virtually never had a case so far were I am confident about my ID that could not be resolved in that way :slightly_smiling_face:

5 Likes

When I download data for research, especially for third parties, I always include the fields
num_identification_agreements
num_identification_disagreements
It is far from ideal, but it gives a good idea of confidence in the ID.

It would be great if one could include whether a specific user added an ID, but short of downloading the data again for each expert in the field (&ident_user_id=123) this does not appear possible.

More at: https://www.inaturalist.org/journal/tonyrebelo/44352-using-inaturalist-data-for-research

1 Like