I feel that duplicates used to be against iNat policy (I guess they still are in the sense that iNat advises no more than one observation of the same individual organism in a 24-hour period). I do agree that some re-education for longtime identifiers might be helpful if the policy on duplicates has changed.
1 Like
I personally have used ‘evidence of a single subject’ to mark as casual landscapes with no clear, identifiable species visible. It seemed a valid use?
2 Likes
From what I understand, the ‘evidence of a single subject’ was introduced to deal with observations that contain multiple pictures, each showing a different species, which might result in pulling in pictures of unrelated species into the CV training set if ID’d just based on the first picture.
5 Likes
Fair enough - and I have used it for that too. But in that case, what is the appropriate way to mark such a landscape photo, which has no valid ID? (Asking for next time, since they seem to like popping up every now and then.)
1 Like
I assume that in most cases an ID of “vascular plants” or similar would be suitable for a landscape. Just because the organisms in the photo are not clear enough to be ID’d specifically and there is no obvious organism of interest does not make the observation invalid.
You can leave a polite comment noting that observations are for individual organisms and suggest that the observer indicate what they were interested in and try to take photos from a bit closer to the subject in the future.
If it is a cityscape without visible lifeforms, or exclusively inorganic material like rocks, water, or sky, you can mark as “no evidence of organism”.
2 Likes
This is specifically mentioned in the new help article
Please do not vote “No” to the “Evidence related to a single subject” DQA condition in these scenarios:
- When there are multiple species in a photo and it’s unclear what the subject is (e.g. a single photo with both a duck and a pigeon in it). A photo with a duck and pigeon and a second photo of just the duck is fine assuming the subject is the duck.(https://help.inaturalist.org/en/support/solutions/articles/151000171680-what-do-i-do-if-the-observation-has-multiple-photos-depicting-different-species-)
3 Likes
When I ID I sort by Date Observed. So I see blurry picture of Is that An Oak, followed by the same, and then seen that already. Twice.
I usually open the obs in a new tab - to check that multiple pictures belong on that obs. If there are Notes, or annotations, or in a project - which all adds up to the info the observer gives us for an ID. I can also see on the banner at the bottom - there is that %%%% oak tree. Again.
If you sort by Random, you don’t see the duplicates and it isn’t a problem to ‘random’ identifiers. I use the Duplicates Observation field. I have no way to take them out of Needs ID. I have abandoned that battle.
Okay, thanks - I must have missed that. I guess also this feels like a different case, where rather than there being specifically multiple species in a photo, there are literally no distinct species that anyone could ID. Yes, it could just be called ‘Plants’, but that leaves it in ‘Needs ID’, and there are plenty of things in there that could actually get useful IDs still without wasting (my opinion…) everyone’s time on observations that can never get anywhere useful.
I haven’t been around for very long (six months or so), so I’m probably still learning the ropes, but I’ve been trying to go through some of the old, languishing high-level IDs to get out of them what can still be got, and ones like this are just frustrating. Maybe I’m too easily frustrated! :-)
2 Likes
Also digging down thru old obs stuck at Order.
You can put in a Feature request for landscape photos. Something that is acceptable to iNat guidelines. There are a whole swathe of photos which capture the wide view for the vegetation type. I had 3 obs for a transect - beginning of line, end of line, middle of line. Valid and useful for the observer - but identifiers need a good way to ID them.
?
For example
https://www.inaturalist.org/projects/vegmaphoto-s-afr
If there are at least 2 IDs and you are convinced that nobody will be able to provide a more specific ID (sometimes even blurry photos may be IDable!), you can check the “ID cannot be improved” box at the bottom of the DQA. If the community taxon is above family, the observation will become casual.
4 Likes
There’s an additional issue related to this, and this is that a reasonable proportion of users will edit their observation in response to requests (removing inapplicable photos, correcting the date, etc.) but they do not communicate that they have done so – with the result that the person who added the DQA vote does not know that they can come back and remove their vote or add an updated ID.
I have started including the request that they should leave a comment once they have edited so that I can update my ID, but many users still seem to edit without commenting.
It also happens fairly frequently that an observer will comment that they have edited the photos but when I open the observation the photos that I asked them to remove are still there. I have not yet been able to figure out what is happening in these cases; I suspect it may be a case of the app not being synced, but without an idea of the cause or how to fix it, it makes following up rather difficult.
There may be an underlying problem here and that is the fact that there is no record that an observation has been edited – not only do we not get notifications that a user has edited the media or observation data, there is also no history indicating that this has happened. This seems rather suboptimal from a data management perspective (changes should be recorded).
In connection with this, it is not ideal that DQA votes continue to be active even if the circumstances under which a vote was added no longer apply (this is particularly a problem for the “ID cannot be improved”, but it also can be an issue for some of the other votes, e.g. “evidence related to a single organism” or “date is accurate”).
9 Likes
I have actually done this recently. A user that I am pretty sure doesn’t have a single observation that he didn’t steal: I managed to find almost all of his bee observations stolen from other iNat users. He has one bee that for all intents and purposes I can’t find anything wrong with, but in a case like this, I think it’s reasonable to assume it is taken from elsewhere as well. This user has 1399 observations, and 9 pages of copyright flags (around 450 observations). It is a lot of work to hunt down where stolen photos come from, so there are cases where just marking it Casual gets it out of the way, as people have done with over 700 of this user’s observations.
In theory I like this, I clean up a lot of Casuals when finishing regions and knowing why something was marked Casual would help sometimes. In practice though, see the aforementioned example of the user with 1400 questionable observations.
I don’t really have an issue with this. We already require observations to have a certain ID precision before they can be Research Grade (Subfamily), why not require a certain area precision before an observation can be RG as well? Especially since if the area is too big, we get into the realm of the observation being unidentifiable.
This gets into the possibility that an observer changed their observation. I have many times told an observer to remove certain photos, only for months later another identifier ask me what I was talking about. The observer had changed photos and hadn’t commented, I didn’t know.
Can’t really do anything about them. But for the record: I hate duplicates. Enormous waste of time.
6 Likes
How do you filter by casual?
I think the crux of it is having DQA votes generate notifications, which, as you say, is already a feature request. As far as other actionable feature requests, I like the idea of having a bit more explanation of the options in the DQA section—maybe not a pop-up while voting, but perhaps an explanation that comes up when you hover over the DQA categories (like iNat currently does with annotation values), or at least a link to the help page that explains in more detail those categories.
I’d be fine with this for any DQA category other than “Organism is Wild”, but making it more onerous to mark something as ‘Captive/Cultivated’ is not something I personally would like (and with the ‘Captive/Cultivated’ box right there on the bottom of identification page, I don’t think iNat does either). I’ll always add a comment when I use the other DQA fields (which I do rarely), but I’m okay with comments being like they are when you add an ID: encouraged, but not required (again, if DQA votes generated notifications, this wouldn’t be as big of an issue).
On the filters tab of the identify page you can check the ‘casual’ box at the top left. On the explore page you can add &quality_grade=casual to the url and uncheck ‘verifiable’ in the filter—don’t know if there is an easier way.
3 Likes
Thanks for clarifying!
And
We understand these sentiments when thinking about identifier-to-observer communication. And this thread seems to highlight that comments may also be read and appreciated by other identifiers. But, again, how often?
But this, too:
And identifiers’ workflows (more IDs vs time for comments)
2 Likes
To be honest, I don’t see how making the user’s observations casual solves the underlying problem; I would argue it obscures it and therefore is a disservice to the community. If a user is repeatedly using copyrighted material for large volumes of observations, it seems like this is a case where more substantial action should be taken further up the food chain by referring it to curators/staff, who have more options for limiting the user’s ability to make future observations or dealing with the user’s observations en masse instead of individually.
6 Likes
I think until there are notifications related to DQA votes, it’s actually quite rude to vote on any DQA without notifying the observer.
In fact, captive/cultivated may be the most important one because we are assessing something that is somewhat subjective, frequently based on incomplete information (which the user may well be able to provide). Furthermore, most IDers don’t seem to have the knowledge to be able to accurately assess whether a plant is cultivated or wild, nor do they have the interest in learning. They usually just vote based on “Is this a native species? It must be wild.” or “Is this an introduced species that isn’t known to be a widespread invasive species (yet)? It must be cultivated.” Some people seem to vote every single non-native plant as cultivated, even widespread invasives, or when they have no idea if the initial ID is even correct. This leads to a ton of completely valid and sometimes scientifically important observations that get buried in casual. I’ve seen entire populations of plants that most people are unaware of because they all get marked cultivated.
Now, will these abusers of the DQA be willing to comment even if this is explained to them? I doubt it, since they’re usually people who are more motivated to get as high on the list of IDers as possible, not to actually improve the quality of data on iNaturalist. So the proposed DQA notifications are very important I think.
But until they arrive I think it should be a part of the iNaturalist culture to always leave a polite comment so a discussion can be had when necessary. It’s probably more common than you think.
2 Likes
I don’t think we should make observations “Casual” just for a large accuracy bubble. I think that in most cases that large “fuzz factor” just means that the observer didn’t know what they were doing with the accuracy. Remember that many people don’t have the slightest idea what 15000 meters (for example) actually is! (Yes, it shows on the map, but it’s really hard to overestimate the cluelessness of some observers. Also, they may see the entire map change color, but not realize that means something.) The observation probably is approximately where the latilong says it is. I think it’s reasonable to assume that’s where it is unless there’s an obvious error.
Note that it’s easy to get a large accuracy bubble if you type in a number, then decide to change it and do so quickly, not realizing the original number is still there. For example, if I decide to change 15 meters to 20 meters, I can easily get 2015, 205, or 1520 meters. (I’ve done this, but [usually?] caught the errors.) With larger numbers, you can easily get a huge apparent fuzz that actually means nothing.
3 Likes
There are some good responses to landscape photos.
- Ignore it. Move on to the next one. Useful for the person who does this, but there are more helpful approaches.
- Ask the observer which species he wants ID’d. Occasionally helpful with recently uploaded photos. This can be an opportunity to recommend that the person take a closer photo. Or that ID might be possible if the photo of this street tree included a close-up of the leaves!
- Pick something in the photo to ID, even if just to Plants or Dicot. With an old photo labeled this way, we should label it “No, it’s as good as it can be,” and it will be gone. I should do this more often.
2 Likes