A tale of a misidentified beetle
Hi Everyone,
One of the great things about iNaturalist is its ability to identify insects using a sophisticated algorithm,. Unfortunately, iNaturalist sometimes makes mistakes, especially when identifying smaller animals (like insects) to the species level. We've recently been going through records of one of the more common Christmas beetle speceis, A . viridiaeneus, which used to occur in very high numbers around Sydney. At first, it looked like there were a reasonable number of A . viridiaeneus sightings being reported on iNat. But after consulting with our Christmas beetle taxonomy expert from the Australian Museum (the amazing Dr Chris Reid) we realised that many sightings may have been misidentified. After reassessment the number of confirmed A . viridiaeneus dropped from 70+ to around 19. Unfortunately, the low number of A . viridiaeneus sightings supports the sad conclusion that this species has likely declined.
This experience highlights the power and the pitfalls of using iNaturalist. On the one hand, it does a reasonably good job of identifying Christmas beetles to genus. In my experience, many beetles get misidentified as 'Christmas beetles' and it would take tremendous effort to remove these from our dataset. iNat's algorithm does this automatically and the amount of time and effort it saves us shouldn't be underestimated. On the other hand, iNat often misidentifies species which can lead to false conclusions if we are not careful (*EDIT* Please see @thebeachcomber's comment below. It seems iNaturalist isn't to blame for incorrect IDs, rather it may be users manually entering the wrong identifications. Making ID mistakes is part of the learning process, so I'm happy to learn that its humans, not the algorithm, making errors. We will perhaps try to put together some training materials to help simplify identifications. Keep up the awesome work everyone!!)
Our team will continue to work through sightings to ensure correct identifications. If we've changed an ID on one of your sightings please do not be disheartened- your sighting is still very important to us.
Identifying insects is difficult and takes practice. iNaturalist helps us to learn together:)
Special thanks to @hauke_koch and @thebeachcomber for their hard work checking beetle IDs.
Cheers,
Tanya
Comments
just to clarify @tardigrade_tanya, none of the erroneous A. viridiaeneus identifiations were from the computer vision/algorithm; this species does not currently have enough observations (even before the corrected records were removed as that species) to enter the training data set for computer vision, so it's literally impossible for it to be offered as a suggestion. So 100% of the erroneous viridiaeneus records were misID'ed by an actual person. You can tell whether an identification was derived from the computer vision suggestions by the little 'luggage tag' symbol; for example, look at my ID at https://www.inaturalist.org/observations/145939120. There's a little tag symbol at the top right of the ID box, between the word 'leading' and the time of ID. If that tag isn't there, the ID was made by a person
also re: "In my experience, many beetles get misidentified as 'Christmas beetles' and it would take tremendous effort to remove these from our dataset. iNat's algorithm does this automatically and the amount of time and effort it saves us shouldn't be underestimated."
--> it isn't the algorithm correcting these erroneous records, it's human identifiers. Users like @hauke_koch and myself have corrected hundreds of june beetles, golden stags, etc., that were misIDed as Christmas beetles by new users.
Thanks for the clarification! I assumed the algorithm was automatically flagging these as A. viridiaeneus as I've noticed it making incorrect suggestions when I use it for other taxa. Thank you for the tip about using the luggage tag- that is a really useful thing to know. I love iNat as a tool but I'm still learning about its strengths/weaknesses.
I have edited the journal post accordingly
no worries Tanya, happy to help; this is a fantastic project!
Add a Comment