I came across an observation today that said it was “submitted with Earthdeck for iPhone” which I’d never seen before. It also had a little link that said “Get Earthdeck for iPhone” which I clicked on and went here: https://apps.apple.com/us/app/earthdeck-ai-wildlife-list/id1592138642
This seems a bit strange to me for a couple reasons. For one, the link feels like advertising (though the Earthdeck app is currently free) and does have tracking tags in it (which I stripped for posting on the forum; they were utm_source=inat&utm_medium=proj_about&utm_campaign=inat). The link is certainly promotional, which I don’t think is supposed to be on iNat (and I found annoying).
It’s also hard to tell how the app works, but it says that it automatically uploads to iNat with one click and “Earthdeck uses an AI-powered wildlife identifier to identify wildlife species in your photos, and automatically organize them in your deck!”
If so, it seems like the IDs submitted are based on some AI (and not the expertise of the user whose account they are being uploaded under). The ID I saw certainly didn’t have the little “AI ID symbol.” That is pretty confusing to me and feels like it violates the spirit (if not the letter) of the rules for identifications.
Does anyone else have experience with this app or observations generated by it? I don’t own a cell phone, so I can’t play around with it myself…
There was another case exactly like this with a different app. That account ended up being suspended, but I don’t remember what it was called, someone else can probably provide more detail.
The app itself (only on iOS) seems to be quite similar to iNat’s Seek, in that it claims to do computer vision IDs on the device and doesn’t upload data unless the user chooses to do that. It’s apparently targeted at adults (whereas Seek’s main focus was children and educational users).
It would be interesting to know what basis is being used for identification suggestions. As @cthawley points out, iNat does have a policy of requiring IDs to have some actual human involvement.
to me, the concept sounds fine. something like this would actually be a nice feature to have in iNaturalist itself, i think.
i don’t think the link was necessarily bad. as a developer, it would probably be helpful to know how people are finding your app. and as a curious iNat user, it’s nice to know what the source of observations may be.
that said, i haven’t actually tried the app. so i can’t really recommend or disparage it.
I’m sure that the link is useful to the developer. However, the source of the observations shows as “Earthdeck” in the observation itself (like where it would say “Seek”), so the link is only really for promotional/tracking purposes.
It seems to get pretty close to the edge (or over) of the Community Guidelines for machine generated content which say that “Machine generated comments would include a generic comment posted via a script.” are prohibited. This technically isn’t a comment (since it’s in the description), but otherwise it meets that definition.
The whole approach also seems close to the edge of these guidelines which also prohibit “Machine generated identifications would include identifications generated from machine learning algorithms or a generic data source with no human moderation/oversight.” Though this would depend on how the app itself works (whether the person needs to manually approve each AI ID being uploaded for them or not).
let’s take another example. here’s an observation from the Wild Orchid Watch Australia project: https://www.inaturalist.org/observations/106783170. if you scroll to the bottom of the observation, you’ll see that the observation was created using the Wild Orchid Watch app, and it links to the Wild Orchid Watch website. would you also consider this borderline promotional?
what about projects that point to organizations like Wild Orchid Watch? promotional or not?
I think there’s an additional promotional component in the Earthdeck case. Observations created by WOW and Earthdeck both have a link to the respective application in the Copyright Info and More section at the very bottom of the observation record. That is something iNat staff has chosen to make available for any app-based submission.
Separate from this inconspicuous link, Earthdeck adds a conspicuous, second, promotional link in the observation’s Notes field:
Observation and ID generated from photo library and submitted with Earthdeck AI for iPhone. Get Earthdeck for iPhone
Personally, I feel that’s not really of value to anyone beyond the developer(s) of Earthdeck and goes against iNat’s policies on promotional content. Essentially, that note is spam (although the observation itself may be just fine). There’s also a question here about whether it matters if Earthdeck is a money-making enterprise as iNat’s current policy appears to focus mostly define spam as content intended to make the author money.
Let’s say the developer(s) of Earthdeck choose(s) to stop adding that Notes text and link. Is there anything else problematic? I think that might depend on how their app suggests IDs for observations that are later submitted to iNat. Has anyone tried it? (No iPhone, so I can’t.) Is there someone from Earthdeck who would like to engage here?
ah, ok. that’s the part that i wasn’t seeing (since there was no example observation provided initially). i looked at one of the observations in the project you noted, and i see what the concern is now. for what it’s worth, the name of the Earthdeck project then (and its description), i would think, would be equally promotional.
i think the project is probably the most problematic thing. the observation note, if it cannot be modified per observation by the user as part of the initial upload, is probably equally problematic. (otherwise, if it can be changed but the user chooses to upload the default note created by the app, i think that’s more gray. i think that would be akin to a footer on an e-mail that indicates which app the sender used to send the e-mail.)
the thing that made Questagame problematic back in the day was not promotional links but the combined account (and therefore the fragmentation of the user base). so i think that’s a different situation there. i would guess that the point of Earthdeck is the automation of finding photos, identifying them, and uploading them. (effectively, it would be a helper tool for uploading.) conceptually, i think this would only be problematic if it does bad stuff like create multiple observations for related photos (instead of allowing the user to combine them into one observation) or loads bad data (ex. a lot of observations without locations) or dissuades users from adding better notes, annotations, etc.
the OAuth app owner and project owner are the same user. so i assume that person would be the primary point of contact.
