perhaps a zoomed out shot, showing the animal at distance, rather than a close up of a paw etc. I think AI training is the job of the photos, and we put the photos, so it follows that we have a significant part to play in the training, whether we choose that monkey or not. We should try and follow the guidelines if at all possible. I’m not suggesting we need police on it, just respect for why they are there :)
2 Likes
Exactly! If we feed the AI pictures that we have told it (through RG community IDs) are examples of organisms, then we are responsible for actively training it to recognize and suggest those same pics as the organisms themselves.
Tangent: That’s also why I like all the feature suggestions of circles/arrows/tagging on a different layer so much…to be visible to human eyes without screwing with the AI.
1 Like
The idea of this being a “sensitivity-slippery-slope” falls on it’s face when you realize the only people who have ever advocated for some kind of warning for dead or gory animals have been the ones concerned about other’s well-being, not some horde of pearl-clutchers.
Besides, there’s nothing wrong with being sensitive. Without sensitivity, why would we want people to connect with nature? This site wouldn’t exist without it.
4 Likes
I pondered a different kind of slippery slope as a posted photos of a dead, mangled dragonfly without a warning.
But anyway. I can’t both try to recruit my botanist friends to post photos and do identifications, and help make this site an unwelcoming place for them by posting mangled mammals without a warning. Sorry, I can’t. I’ll see what I can do about posting a text warning on a partly obscured photo showing only the least offensive part of the dead animal.
iNaturalist has thousands of observers and identifiers and users of data, all with our own goals. Training AI is a good goal, but it’s not mine. (I think of the photos I may post for a grass – ligule, whole plant, spikelet, shot through the leaf to show some feature visible by transmitted light, inflorescence, etc. – and think, if the AI trainers can deal with all that, surely they can help AI ignore black-on-white text. Surely.)
2 Likes
Computer vision training is only one aspect. If I’m looking up images of an organism I don’t want to see a bunch of irrelevant “warning” graphics.
10 Likes
Yes, there is that.
2 Likes
As part of evaluating their approach to handling this subject, Tony is looking for people who have genuine concerns over distressing content. While you might not have a level of distress yourself, if you know others that do, perhaps you could help Tony with your experiences. Certainly, if you feel strongly enough about it, you should continue to advocate in that direction. I would suggest direct message him and share with him your experiences. For me personally, I would support features that allow others to control what level of distressing content they might encounter, but it is at a conceptual level, as I don’t actually know of anyone who has been distressed to that extent.
1 Like
I’m distressed by images of plants that have been clipped, disarticulated, and squashed in a plant press. But maybe that’s just me.
If you don’t hear from real iNat users who are actually distressed by roadkill and similar pics, is this really an issue that needs a solution?
4 Likes
When I’ve heard feedback about this at help@inaturalist.org, it’s been due to a few main concerns:
- belief that the observer has purposely killed/mutilated the roadkill animal.
- belief that showing dead/roadkill animals goes against iNaturalist’s goal of encouraging conservation.
- when showing off nearby observations functionality in a presentation, or to friends, a bunch of roadkill animals sometimes pop up.
These are all understandable, and I think the first two concerns are quite easily allayed by an explanation about the importance of this type of data, and that it’s highly unlikely the observer killed the animal. In my experience that works and people are pretty interested in how roadkill observations are important.
The third concern is also understandable, although I’m not sure how common it is. I think an alive/dead annotation plus preference toggle could reduce the likelihood of this occurring to almost nil.
5 Likes
It’s not about dead OR alive OR birth. The topic is: are some things too gory to include?
I bet more people would consider an image of mammalian birth more gory that of a dead animal. We have lots of kids newly exposed to nature here. Do they really need to learn HERE that birth involves so much gore? I can’t make up my mind about that one.
Awkwardly, “too gory” is different for everyone.
P.S. Will anyone be upset about a photo of the mosquito I swatted as she bit me? And can someone who knows mosquitos please check it out 
:
https://www.inaturalist.org/observations/25487987
1 Like
Well, kids under 13 on here should be supervised by adults and when 13+, given they are mammals whom could become pregnant or cause someone else to be, they probably SHOULD know what it entails. Not that iNat needs to be the place they learn about these things but I don’t see hiding the reality of childbirth from teenagers as a worthwhile goal of inat either.
7 Likes
Yes, I know. I was specifically replying to jokkomarat’s comment:
I was clarifying that I was presenting an alternate example of entrails, to contrast with the bloody ones, and asking whether both would be considered equally “gory” under the warning systems people were proposing.
1 Like
My point was that WE are the AI trainers. ; )
We train it by IDing observations.
Which to me, is part of the larger issue Cassi pointed out about not IDing warning text as representing the organisms.
I think I’d be okay with block of text that instead of being a generic warning actually describes the content of the rest of the pictures in the observation, though. I wouldn’t consider that functionally too dissimilar to the currently allowed diagnostic non-photos in the “photo slots”…things like drawings, or seeing the graphical depiction of a sound wave without an audio file (some of the bat observations do that). As long as it is not something really vague like “viscera in the following pics”.
The innocuous zoom is probably the best option, though, and of course the preceding opinion on the type of text is only my own ramblings.
1 Like
It’s worthwhile coming up with ideas before it becomes a problem. Look at how long it takes to come to anything even approaching consensus. Let’s get the talking out of the way now so that if and/or when it becomes an issue we don’t have to waste time debating.
1 Like
To complicate the matter further, whether something is really disgusting and unpleasant to see doesn’t, in my opinion, depend simply on whether it’s dead or not. Today, for example, a neatly dead road-kill Striped Skunk was posted. Not at all offensive, I think. A few weeks ago, though, I posted (with warning sign) a seriously mangled road-kill rodent. Offensive.
Apparently it would be good to respond to tiwane, who wrote:
I found this statement puzzling when I first read it, and I still do. You might be concerned about people being offended, but not concerned about them being disgusted, sickened, or hurt by the photo? Seems backwards. But perhaps it’s because of different ideas you and I hold about the meaning of the word “offensive.” To me, someone who is offended by something he/she thinks is wrong and is angry about that. I figure being polite involves minimizing cases where the other person will be offended, but being careful not to cause disgust or pain is at least as important. (If all we have here is a difference of terminology, let’s just call it good and go on.)
I worry about times people are disgusted, hurt, sickened by the photos. And disgust may not have the implications you seem to assume.
Example involving me: A series of photos of dead, smashed, road-kill newts on iNaturalist. And soon after it ended, another series. And a couple days later, another series. I totally agree that those photos document something important (perhaps a migration point that takes newts across a road where they get killed). I totally agree they should be posted on iNaturalist. I also found the cumulative effect disgusting, somewhat sickening, and I strongly wish I could have avoided seeing them (without just avoiding iNaturalist!). I wouldn’t say I was offended by them, but I found them somewhat hurtful. (Perhaps I should also conclude that I’ve spent too much time on iNaturalist this year.)
I also think of a field project in which I spent an evening slowly soaking and cutting small bits of dried skin off a coyote skull because my botanist colleague wanted the skull but found the bits of dried skin too disgusting. I know he can deal with disgusting things when he has to, but his feelings about these things are strong and he avoids them when he can.
Therefore, based on n=2 (or 3, as I think of another botanist colleague, or maybe 5, considering relatives well known to faint at the sight of blood), I extrapolate that there are some people out there who will find looking at photos of dead, mangled vertebrates disgusting, painful, hurtful, and I want to respect those feelings while also posting the disgusting photos. I will try to do so. That will mean not posting an obviously mangled mammal as the first photo, showing it to unprepared people. It may mean posting first some explicit warning like “This observation shows the disgusting, mangled dead body of a deer that was hit by a car here.” It may mean combining the warning with a cropped photo small bit of fur or hoof so I meet the (entirely reasonable, most of the time) iNaturalist requirement that some part of the real thing be included in each photo of an observation. But do I feel a need to do something about these photos.
5 Likes
I think the difference is in how you would react to it. For instance, if roadkill images disturb you, then you might respond by being more proactive at not hitting animals while driving (slowing for corners or near animals by the side of the road etc). If they offend you, then you might quit using iNat.
google definition of the two:
disturbing
/dɪˈstəːbɪŋ/ *adjective* 1. causing anxiety; worrying.`
offensive
/əˈfɛnsɪv/ *adjective* 1. causing someone to feel resentful, upset, or annoyed. synonyms: insulting, rude, derogatory, disrespectful, hurtful, wounding, abusive, objectionable, displeasing, annoying, exasperating, irritating, vexing, galling, provocative, provoking, humiliating, impertinent, impudent, insolent, personal, discourteous, uncivil, impolite, unmannerly, unacceptable, shocking, scandalous, outrageous * (of a sight or smell) disgusting; repulsive. synonyms: unpleasant, disagreeable, nasty, distasteful, displeasing, objectionable, off-putting, uninviting, awful, terrible, dreadful, frightful, obnoxious, abominable, disgusting, repulsive, repellent, repugnant, revolting, abhorrent, loathsome, hateful, detestable, execrable, odious, vile, foul, unsavoury, unpalatable, sickening, nauseating, nauseous, ugly, unsightly
Looking at the two meanings, it’s hard to tell the difference in the context of a roadkill observation, but I would be more worried about creating anxiety in someone else than I would of being rude to them. I guess as someone who suffers anxiety I therefore place more emphasis on the negative effects of that.
4 Likes
If warning signage is an issue, in part because it might mess with the computer vision learning for that particular taxon, then showing parts of an animal (e.g., internal organs) could also do that. So can including images of animal sign such as tracks, scat, beaver dams and lodges, etc. where the animal itself is absent … unless the CV is capable of learning to recognize this as sign and not the animal.
1 Like
Minor clarification: observations just require evidence of the organism, which may be the organism itself, or tracks, scat, dams, nests, shells, etc, which means that the computer vision model gets trained on identifying photos all of the above.
6 Likes
Possibly … given some scale in the image and an idea of location (narrows the options down) it might be possible. (I was a pathologist before retirement). An interesting challenge anyway.
1 Like