The benefits and drawbacks of adding coarse identifications

[jeanphilippeb]

@sedgequeen

(1) it’s easy enough to prevent that loss by copying the placeholder into the comment when I add an ID and (2) this is something iNaturalist really should fix.

(1a) New iNatters, and possibly many iNatters in general, don’t know that.

(1b) The “Identify” page (showing 30 observations) does not even show if there is a placeholder.

(2a) iNat didn’t plan to change the placeholder:
https://forum.inaturalist.org/t/question-about-placeholder-notes/43210/3

(2b) Still waiting for this other “should be done” (unless I missed it) (unrelated topic, just an example):
https://forum.inaturalist.org/t/implement-a-new-filter-for-observations-that-opt-out-of-community-id/27053/18

I enjoyed doing things, as I could, instead of just waiting (for how long?) for new features to come.

[DianaStuder]

Because your list of possible birds is short.
Compared to all the plants in the habitat around the birds.

@environ

keeping those obs off the screens of people who only care about ID’ing Animals

that is the intention tho. Take all this green stuff away.

Insects are another group where many are not yet described.
I can fish - and the ID moves.

Yellow label projects use iNat CV to pre-select what CV would offer.
It is a different way of tackling Needs ID.

[jeanphilippeb]

@environ

And if I gather correctly from your journal page, if you’ve updated those projects recently?

A software running on my computer updates these projects everyday (and for a few months another person helps me running this software).

Limitation: 10,000 requests to the iNat API per day (requests to the iNat server).
So, by now, with 2 persons, 20,000 requests/day.

1 request to get the computer vision suggestions for 1 observation.
1 request to push 1 observation to 1 project.

[sedgequeen]

I see now! You’re trying to direct me to the Yellow Square projects! They’re great! Sometimes I use them. Then I forget what they’re named. I haven’t forgotten Poales (yet), though, so revert to sorting on that or other taxonomic names. (I often do the most general IDing when I’m tired and don’t have much patience for searching through for a forgotten project name.) Anyway, thanks for setting these projects up! They really can be a help.

[jeanphilippeb]

@environ

it looks interesting, but not exactly user-friendly yet

Traditional projects are the basis of this work, I use them and I don’t know a better tool on iNat to get a similar result. You can browse these projects from whatever project you start with, following the links to the parents projects or to the sub-projects, in the “About” paragraph of the project.

Each of these projects has direct links for identifiying. I never create manually these URLs, I just pick up the links from the projects. (Of course it would not be user friendly to write these URLs. There were all generated by a software). That should not be an issue once you know what you are interested in and find the corresponding projects.

Yet, it is up to you to change the filter (click on the “Filter” button), for instance if you want to exclude the “Casual”. I need not explain that.

There is also the journal post with the whole list of projects, displayed as a taxonomic tree, for an overview and a direct access.

[jeanphilippeb]

@spiphany

For better or for worse, most IDers are going to be sorting based on the ID attached to the observation. For this to happen, they need broad IDs.

Specifying the solution need after expressing the need solution is a considerable bias in the vision. The need comes first, the solution comes next.

We need to take a step back from this vision.

[Nothing personal in my response. I am aware that I took this citation out of the context. But, for me, this popped up so clearly!]

[DianaStuder]

@spiphany

most IDers are going to be sorting based on the ID attached to the observation

Your ‘most’ would be only taxon specialists with filters set.

I suspect the actual most identifiers are split between - dipping in, don’t care, moved on - and plodders like me trying to make headway.

[wendyjegla]

Ah, this is good to know. I’ve wondered about this kind of issue with regard to Hymenoptera. E.g. for a given observation, CV suggests a family in Ichneumonidae. I do some due diligence and I think it looks like a reasonable guess. But there’s nothing reasonable about Ichneumonids - or Braconids - and the CV and I could be wildly wrong. The best thing for me to do in this case is whatever is best for the experts. Would they prefer I go with the safest, broadest guess, i.e. Ichneumonidae… or put it in a more narrow bucket? I would think that trying to go through Ichneumonidae observations is a lot harder on one’s brain than working on a subfamily or genus set.

I figure this is probably a matter of preference, and it’s hard for me to anticipate what’s preferred. So whether I put an observation in Ichneumonidae or something smaller ends up being a matter of how brave I feel that day. But it sounds like judicious use of more narrow taxa would be the way to go…

(Given all the discussions about taxa-guessing, you’d think I’d have figured out the right thing to do by now!)

[ItsMeLucy]

Just a note that the best course of action could vary by geography. There are definitely Ichy peeps who cull through regularly specifically here, as well as bee verifiers, a local beetle person, butterfly guys, etc. But this may not be true in other areas.

Then there people who have more specific pet groups but worldwide; I know if I post any member of Stratiomyidae who is likely to catch it and within what timeframe (about 24-48 hours because he is wildly attentive).

[psweet]

Wholeheartedly agree – as a researcher, it’s your responsibility to know the limitations of whatever datasets you use, even if someone else collected them.

[lappelbaum]

@jeanphilippeb

Yet, it is up to you to change the filter (click on the “Filter” button), for instance if you want to exclude the “Casual”. I need not explain that.

That reminds me I meant to ask. With the yellow projects, when I open observations then click on Identify on Filters it always defaults to Quality Grade “Casual” instead of “Needs ID”. Why? I never ID casual stuff so it is not sticky from something I was doing earlier.

[wendyjegla]

Oh, good point! You know, what we need is a gigantic grant to fund salaries for full-time ID staff.

[environ]

@wendyjegla

So whether I put an observation in Ichneumonidae or something smaller ends up being a matter of how brave I feel that day.

If you’re making the first ID, or an improving ID for some observation - then I would say Be Bold and be as specific as you think you reasonably are able to be, even if you are not Stake Your Life On It certain that’s correct - because that gives it the best chance to be seen by whoever self-appoints themselves as an expert on what your best initial guess might be (and if they disagree, you get a chance to ask them why and learn a bit more).

But if someone else has already ID’d it to your best guess, then I wouldn’t add a confirming ID unless I was very confident that was correct, because now you’re voting for that ID, not just suggesting it as something people might consider.

[sbrobeson]

@sedgequeen

They do have that option and they should use it!! Too many researchers assume that RG observations are all correct and then complain when they find that they’re not. Researchers should at least check the outliers in their datasets!

@psweet

Wholeheartedly agree – as a researcher, it’s your responsibility to know the limitations of whatever datasets you use, even if someone else collected them.

unfortunately it’s not always about outliers, but often about what I can only think to call “inliers” – incorrect Research Grade observations being exported to GBIF, where the issue is (for example) cultivated records left by default marked as wild, or misidentified records, which can really destroy things like niche models not necessarily because they’re beyond the natural range edges, but because they’re probably in habitats with environmental variable values that distort the “habitat preferences” or niche envelope that’s to be inferred. there’s no way to automatically filter these out… it requires going through the records manually, in many cases thousands of such records. most researchers don’t have time to do this, and to be frank, most researchers don’t even understand the problems with iNaturalist-sourced data. many of my colleagues didn’t even know iNaturalist data went into major sources like GBIF! they thought they were signing up for a rather different set of limitations than they ended up getting. and while GBIF thankfully allows for deselecting iNaturalist data entirely, if one does choose to accept this data source, then no amount of filtering will ever quite get to the point of looking at every observation one by one, especially for the most common problem taxa (such as any cultivated tree ever). I do this for my focal plants, because the total number of observations is small – in the low dozens of hundreds, maybe. someone researching Quercus macrocarpa (and I know several such researchers!) won’t have that luxury given the volume and time involved. I agree on the principle – know for what your data are, or are not, suitable. it’s just that iNaturalist data tend to be limited in ways that are only clear to those with an intimate familiarity with how iNaturalist itself works, and that’s not something you can assume scientists will be able to do from a basic reading of the site structure. the amount of explaining I’ve had to do for non-iNaturalist-power-user colleagues is discouraging, to them if not to me too.
all of my examples are just examples – there are probably other problems I haven’t thought of in the span of 5 minutes.

[environ]

@sbrobeson

it requires going through the records manually, in many cases thousands of such records. most researchers don’t have time to do this, and to be frank, most researchers don’t even understand the problems with iNaturalist externally[ed.]-sourced data

…and that kids, is what’s wrong with most of the published papers in this week’s Esteemed Peer Reviewed Journal!

But more seriously, if you are blindly assuming things to be absolutely true about any data set which hasn’t been confirmed by careful analysis, then all of your conclusions from other analysis of it are going to be suspect.

And maybe that’s fine if what you are publishing is just grist for the treadmill and giving someone else something to analyse and refute later - but Garbage In, Garbage Out is a pretty fundamental principle and applies as a very real caveat to any data you didn’t collect and carefully inspect yourself.

– There’s never enough time to do it right, but there’s always time to do it again.

[Mecynogea_palustris]

• People prefer to have their ID’s confirmed or denied in shorter time periods.
• I’m unsure if this is a projection, but I feel most Identifiers filter by family, or higher taxonomic levels.
• Therefore, ID’s with higher taxonomic levels, ie genus or even species, tend to get more attention than others ID’d at lower taxonomic levels, ie phylum and order.
• This means there is an increased incentive for users to ID their own observations at higher taxonomic levels regardless if they are incorrect.

[Mecynogea_palustris]

There’s so many misidentifications of obsucre taxa like Alcyonidium and Sarcodiotheca across the mid atlantic that the CV now promotes more people to ID to that speices in ranges where it is not found.

[environ]

If you know those exist, that should be one of the easier curation jobs to do a sweep for - you can just Explore those regions for them and pretty quickly add disagreeing ID’s for them with a copypasta explanation about being out of range.

It might even be something you could solicit help for if there are vast numbers of them, with a post to the forum explaining the situation, how to find them, options for what to replace the prior ID with - and a link to that topic with each disagreeing ID to keep all the discussion in one place for everyone to see.

There would be a notable proportion of my disagreeing ID’s that fall into this category - where I’ve found someone has ID’d a lookalike from another region.

[Mecynogea_palustris]

I’m not an authority on these subjects, (I work IT not real science) so i rarely feel comfortable enough to correct entire regions of people based on a paper I read in my freetime.

[ilenisaatio]

I’ve never understood why there’s a need for CV at all. At least one that goes into any more specific level.

My personal observation concerning the topic is, that there are so many cases of me going through unknowns from over a year ago and adding even a broad, sometimes a kingdom-level id, and then there’s a sudden influx of IDs diving down to expert IDer.

I’m aware this is just anecdotal at best, as I’m not the person who uses the data or has deeper knowledge of the inner workings of the platform.

But I stress again, that this is a citizen science project, and a complex one at that, so the quality of data is always suspect. It’s how these things work. Things like this give a rough sieve. If there’s an issue with the data going to GBIF, then take it up with them or the heads of the whole iNat-project. If you need a clearer and more obvious disclaimer for people who intend to use this data, do the same. I’m pretty sure most people here are doing this as a hobby and not a furrow-browed semi-profession.

I can’t understand how experts and professionals keep getting this wrong. There are cases where citizen science produced data can be used, and there are cases where it can’t.

Michel Mayor of University of Geneva, and Emma Lundberg of The KTH Royal Institute of Technology seemed to be happy with the results they got from their respective citizen science projects they arranged through the MMORPG Eve Online, but they used it to do preparatory work, and had their own teams go through the consensus results separately.

If what @jeanphilippeb and @sbrobeson say is true, then there is an issue, I agree, but it is mainly an issue of misunderstanding the function of projects of this kind in my view. You need source criticism, but if citizen science projects are opt-out, then that should be rectified to opt-in IMO, and if you include them, you should get a clear disclaimer that notifies of the limitations.