Platform(s), such as mobile, website, API, other: Desktop, optionally apps
Description of need:
After a) reading about/from frustrated identifiers and b) my personal experience uploading:
Change the order of species/date/place to place/date/species. That way more people would naturally fill in the place first and then the computer vision is more precise, not suggesting things from another continent.
Naturally, we fill the top field first. To fill in species first make sense if people know what they are looking at. However, a big amount of iNat-data is not uploaded by observers who are not sure what exactly they see.
New users would profit from being guided to make the most out of their observations and this way get more often a correct identification that is the case now
For experienced users, it’s more a personal preference I would say. However, I personally think usually where it was first, so it does feel more natural like that as well.
Here are some examples of how that would help inexperienced users who might just click on the first options:
The comments so far seem to assume that users are all submitting observations. In my experience, many users enter the photo, look for an ID, and then say “oh that’s what it is” and back out without submitting anything. If they had to “fill in” a place or time, they wouldn’t use iNat at all. Now, I’m not sure if that’s a good thing or a bad thing! I do think putting place first would make the whole experience clunkier, but let’s see what others think.
I don’t manually enter a location since the GPS coordinates are embedded in the image. I don’t know how many users are in the same boat, but to me, shuffling around the input fields would be an annoying change, for no apparent reason.
If the goal is to give better suggestions via the algorithm, then coerce the user to provide a location before giving any suggestions. Sorry, I don’t know how to do that without annoying that user as well.
Not in favor of this. My primary data entry and research efforts are species-oriented, not location oriented. I can always modify the location if I want to set it, narrow it, broaden it, etc., but my uploads and my inquiries are almost always taxon-based, so I’d rather not see those entry boxes switched.
It’s true that I wasn’t considering users that are not submitting observations, but just using the upload display to try to identify organisms for their own private purposes. But I think making that usage slightly clunkier is well worth any improvement to the set of actual observations.
I do think it would catch the attention of existing users, and probably annoy some of them mildly at first (as any software change always does). But it doesn’t make it any more difficult to upload photos with embedded locations, so I think people would get used to it quickly.
If someone is to be annoyed, I think it should be the user that is making suboptimal use of the algorithm. Arbitrarily altering the user interface in hopes that that user will change his/her behavior is probably wishful thinking. Even if the input fields were shuffled, the user might go straight to the algorithm. We’ve been trained to do that.
What I see as the upside to this change is that iNaturalist would get better initial IDs, which seems like it helps everyone involved. I don’t think this would be arbitrarily altering the user interface; I think it would be altering the user interface to achieve better data, with the assumption that most people fill out forms from top to bottom. I think this assumption is almost certainly true.
I see this part of the current interface design as rather arbitrary. Users expect to complete input fields from top to bottom, but iNat’s logic is to use data from the location and date fields to inform species name suggestions in the box above those fields.
iNat is about people sharing what they have seen in nature and trying to correctly identify it. Deciding on a taxon is the output of that identification process, so it makes sense to ensure there’s a good location and date before iNat tries to assist users by suggesting IDs.
This change will affect only a small portion of observations (those uploaded via the web interface where the first photo lacks location metadata or a time stamp). But I’m pretty sure that the path to improved ID quality within iNat lies through the accumulation of many small UI tweaks and education steps.
A lot of us upload photos from cameras rather than mobile devices, and we either don’t have a GPS in the camera, or leave it disabled to save battery space. This means that you have to manually enter the location. Generally I do it via the map as it’s faster and I can more accurately get to the exact location.
Even when you’re uploading photos from a phone, you’ll want to check that the coordinates are accurate. I ran into that while sorting photos for upload from The Doctor’s iPhone: two photos, in the same sequence, taken seconds apart, had wildly different GPS data – more than 10km apart. (Thankfully, one of them was correct, so I didn’t have to fill it in manually.)
My camera eats batteries fast enough as it is; I can’t imagine what having onboard GPS would do to my supply!
Just to clarify, I don’t think that the request is to force users to enter a place name first, just to move the place field to the top place. There’s nothing to stop a user from clicking slightly lower and using the exact same workflow with the species box that they currently do. The proposal as I read it wouldn’t add time to just uploading a photo and getting an ID without giving a place, just require a small change in muscle memory.
I don’t think that this change would be a burden for experienced iNat users who know how they want to interact with the site. But for newer, less experienced users it would be a very gentle guide to improving observation quality (and potentially saving a lot of identifier work). Currently, identifiers can’t keep up with the flow of observations to ID, so any small tweaks we can make in improving identifications could help the overall user experience imo.
I find the Computer Vision Demo much better for that specific use-case. The layout of the page is much simpler (with a big green Classify! button), and the suggestions show a gallery of six large thumbnails per taxon, instead of one tiny thumbnail in a drop-down list. The demo seems to load the metadata a lot more quickly than the uploader, too.
My phone (the major source for my photos) does not add GPS coordinates to a photo. I don’t think adding a location is a big deal because when you click on location the map comes up and you can find your location fairly easily. I do think the suggestion would be a win-win because of the impact on the AI, and the training effect of “Fill in this first box first.” The training effect might not work very well (but could be tested). But still, the improved AI with the input of location would still be a win.