Beginner iNatter here (and beginner with observing species).
Today I was trying to upload a fly. The AI pictures did not look right at all. However, when I tried looking at other flies that weren’t found in the area (or even necessarily on the same continent), they looked more similar.
Should I just identify it more generally as a fly and hope someone helps i.d. it? Or should I select the particular species that I think it looks like, even if it’s an incorrect identification, where someone who confirms i.d.s can correct it?
I don’t know how annoying that is for those confirming i.d.s and as they are already working hard enough, I wish to use the proper etiquette. Any advice would be greatly appreciated.
From my understanding, it’s preferable to give an ID to your observations rather than leaving them as “unknown”. Observations with an ID are generally more likely to be seen by identifiers than observations which do not.
So in your case, I say go ahead and add the ID “Flies”.
If you do choose to go with the Computer Vision less curtain suggestions, be ready to withdraw the ID if it turns out to be inaccurate. It can stall the community ID if you don’t.
Go to a level you’re comfortable with. Even a high level ID is useful as it will get the observation in front of people who have filters set up for certain types of critter. Also bear in mind that there’s no harm in getting it wrong. The iNat hivemind fixes most things.
Flies in general are very diverse and difficult to ID from photos. The AI is not good with them, aside from a some very distinctive groups/species. Tachinidae, Muscidae, Sarcophagidae (things that generally look like “house flies”) are especially difficult so don’t be upset if it doesn’t even get to genus. There are people who triage flies though - moving them from Diptera into whatever family/genus they can. And some experts who periodically review from there.
Usually when I have an unknown fly that I can’t reasonably place into a more specific group (like fruit flies, picture-winged flies, hoverflies, etc.) I just leave it at Diptera and hope someone eventually sees it. In this case it probably doesn’t matter too much whether you put “Flies” or a particular species, as long as you put some kind of ID on it.
PS: check out @edanko’s Fly Guide, may help you get a little closer than “Flies”
Even if you’re not a fly person (I’m not), you can usually split them into Nematocera (long, often brushlike antennae, overall look is usually mosquito-like) vs. Brachycera (short antennae, overall look is usually house fly or fruit fly-like)
It is always better to provide an ID to your observations. But when you provide the ID, I feel it is always best to go to a classification level you are comfortable with.
Thats good that you are at least critical of the computer vision, and not just blindly rely on it. I recall from a thread here that the computer vision takes into account the initial guess, so if that is specific and incorrect then it can suffer (although now I’m not sure if that is the case), so that is why I personally prefer to try to be as specific as possible without providing the wrong ID.
If the CV gives a suggestion that’s similar but clearly wrong, take a look at the taxonomy. For example with with a Common House Fly (Musca domestica), there are a few tabs you can select, one of which is Taxonomy and looks like this:
You can go up the taxonomy tree selecting what group you think it best fits into. This is helpful even if it’s something very broad like Animals because it’s still narrower than unknown and would show up for identifiers who only look there, ignoring out Plants, Fungi, and other kingdoms, for example.
It’s perfectly okay to be wrong with an identification. If someone suggests a different ID, maybe ask them why they suggested that over the one you selected? I think this is just the normal learning process, nothing annoying about it. I’m sure the folks who have 1000s of IDs have probably misidentified at least 10s of species before.
I enjoy the identifications where I’m wrong or the observer is wrong and there’s some discussion about why more than observations where I’m just agrering with an ID. Not sure how true this is for people who identify 100s of observations/day, though.
It took me a long time to get a good idea of how exactly I should be identifying things, and I wish I had known this earlier:
Don’t add an ID unless you are reasonably sure that it is correct.
It’s good to be bold, and make a suggestion if you think you know, but don’t just guess at it. The CV suggestions are getting better with every update, but they’re often still terrible and if nothing looks right, then none of them are probably right!
It’s much much better to add a correct but broad ID than to guess and be wrong. This is especially true in groups that don’t have many identifiers, as there may not be anyone to correct your mistake. Even a really broad ID is helpful. E.g. I know virtually nothing about plants, so when I upload sightings of plants I almost always just ID them as “Angiosperms”. That’s the extent of my plant knowledge, so that’s what I’m comfortable IDing.
Sometimes it is okay to guess a little, but you’ll get the hang of those sorts of situations as you start using iNat. In general, you should note when you’re not completely confident with an ID in the comments, and always be open to other people’s thoughts and corrections.
Two more things: Firstly, you should always add an ID to your own sightings, but it doesn’t have to be precise - as I was saying with my plants, just a very broad ID will get it partway to the people who can ID it. Secondly, you shouldn’t agree with someone else’s ID just because they’re an expert - an ID should be something that you can verify yourself, even if you couldn’t have come up with it by yourself. E.g. if a world-renowned orchid expert comes and IDs one of my orchids, I won’t agree with them just because they’re an expert. But on the other hand, if they ID mine and then comment something like “this is the only species in the area with purple flowers”, then I can go and look that up, see that it’s the only species with purple flowers, and ID it, because I was able to verify it myself rather than just blindly trusting the ID.
Anyway, hope this helps, and welcome to the iNat community :)
Keep your ID at a broader taxon that you are confident of. If you add a tentative species and someone agrees with you … dipterist has to go thru the research grade obs, check each one if it is correct. If the dipterist disagrees with the previous 2 IDs then iNat needs the dipterist and 4 more to counter the 2 which prove wrong. Even if you withdraw and there is only 1 wrong, iNat needs dipterist plus support from 2 more to tip the ‘over 2 thirds’ balance.
If you put up a tentative ID you need to respond promptly to your notifications. That works to get IDs moving! But Diptera plus patience will get you there too.
That also what I am doing to my plant observations. But some of them remain unidentified for a long time. I think not everyone check large taxon for anytime. So I have to @ my friend who know plant better to help.
In a case like this, you could check, say, the three species in that list that look most similar. If they are all in the same family/superfamily/subfamily, it would be reasonable to add that family/superfamily/subfamily as a tentative ID. But still be ready to withdraw that ID if others then vote against it. If they’re not even in the same superfamily, you might need to go up a level, to infraorder or suborder, as @bugbaer suggested.
I agree it’s a good idea to leave it vague if you are unsure. However, I would argue it can also be beneficial to be bold and dare to pick a species, even if you are not 100% certain it’s correct. It should be an educated guess after doing some research and your best approximation at what you think is most likely and not a random, wild guess. You can put a note on the observation that you think this is the most likely ID and explain why you think so but you’re not fully sure. That alerts others that they shouldn’t just click agree without taking a closer look.
Of course it’s best for every ID to be correct to start with, but even if you got it wrong there are still benefits I think: 1) It may attract the attention of experts who are following specific taxa but may not look at observations ID’d to higher levels. If that happens and they correct you, that’s a great opportunity to ask and learn. 2) Even if initially wrong and it gets corrected, it will pop up in the “similar species” tab on the taxon page until it goes to RG. Some identifiers (myself included) will occasional look through those as an indicator for what taxa are worth checking IDs on to correct wrong IDs and improve CV suggestions.
I’d suggest that if you’re going to guess, even an educated guess, it is better to put the ID at genus level and mention “maybe species X?” as a comment with it. Otherwise, what do you do when someone just “agrees” and it goes to Research Grade? You won’t know (unless you can see from their profile they know about the group) whether they’ve brought any additional expertise to bear. There are a number of identifiers (not many, in fairness, but they are sometimes high-volume when they appear) who go through agreeing to any ID, including wrong ones.
I actually have done this before… actually even just yesterday. Beeing a bit bold… but then also tagging someone I know knows this matter better then me in that area and getting into an discussion about it (the boldness from yesterday turned out to be wrong btw … but it was very interesting for me to be able to learn from someone more skilled… and retract my ID).
I agree that it might not be adviceable to be bold just “out in the void” and hope for the right person to see it… if you ever decide to leave that void, it can be very difficult to overcome a possibly wrong ID by the ones remaining.
If you are deciding to use visual recognition tools to ID it’s most reliable to do so in Compare or Suggestions vs. clicking on the ID search bar. There you can use the visually similar filter and also check all ID suggestions for the location. It also helps to broaden the location filter (check county, state, country, global/no location) and the taxonomic filter (check subgenus, genus, tribe, family, order). And to consider whether some ID options have never been observed/IDed yet, since if so they won’t show up unless someone manually added them to inat place-checklists. Experts often suggest erring on the side of caution with species vs. coarser IDs, or to indicate an ID is uncertain in comments. It’s also more consequential to make a secondary confirmation of another user’s species ID than the first ID, since confirming makes the record Research Grade.