The current Android beta has some much-requested changes to the taxon chooser that we’d like help testing:
Only nearby visually-similar suggestions are shown by default
You can choose which photo in an observation to generate suggestions for
Filtering results by taxon
Ability to only show suggestions from RG observations (i.e. only nearby regardless of visual similarity)
Some specific things we could use help testing:
How’s this working on older phones, or older versions of Android?
Does the text make sense, e.g. “View suggestions not seen nearby”? We debated this a lot, and our compromise solution is… ok, but could probably be improved
Can you get it to crash? If so, please provide detailed steps to replicate
Switching to Research Grade suggestions instead of Visually Similar doesn’t work, sometimes closing the app but mostly just exiting out of the photo viewer. (Sent logs.)
I’m not super sure what the RG search would do. Allows you to search for nearby RG taxa instead of visually similar taxa? I would call that Nearby instead of RG. Or Nearby RG.
What’s the reasoning for the binoculars icon for the Suggestions Source? Not really complaining but it’s the same one used for the observations tab on desktop.
Selecting the compare button on a suggested taxon compares it with the first observation photo, even if a different photo provided the suggestion. (Not too hard to scroll over though.)
Protozoa is not available as a taxon filter dropdown option as it is on desktop Explore. (But you can still type it in.)
I think “View suggestions not seen nearby” makes sense. Maybe there’s a better way to phrase it though.
The “We’re pretty sure…” can change if you’re restricting to different taxa. Maybe that’s supposed to happen, but I suppose it would be “iI you’re pretty sure it’s … we’re pretty sure it’s …” And the taxon filter should probably go above the “We’re pretty sure…”.
** “View suggestions not seen nearby”** = View all suggestions
** “View suggestions not seen nearby”** = all suggestions-** “suggestions seen nearby”**?
I’m not exactly sure what you’re asking but it looks like selecting “View suggestions not seen nearby” adds AI suggestions that are visually similar but have not yet been observed nearby, in addition to those that have been seen nearby. It makes the suggestions the same as what they were before the update.
Maybe “Add suggestions not seen nearby” would be better because otherwise it sounds like you’re opting to only view suggestions that haven’t been seen nearby. It looks like it just added in however many options would add up to 10. Not sure how they were ranked though: the suggestions that were also seen nearby were interspersed with those that were not also seen nearby.
Is this only happening for an obs that has no coordinates and/or no existing taxon chosen? The major problem with suggestions from RG obs is that without any filters… that’s often a lot of suggestions, which is a very large API response, which might cause a crash if your phone doesn’t have a lot of free memory. We can (and maybe) should cap the list at 10 like we do for visual suggestions, but that kinds of defeats the purpose of the feature for me, which is see everything that people have observed within a specific taxon & place query.
It should show you leaf taxa represented in RG observations given place and/or taxon filters, e.g. “show me all the maples observed in New Hampshire” or “show me all members of the Plethodon”. It’s not necessarily just nearby stuff.
It’s an icon we use to represent observations… sometimes.
Hm. The Compare tool compares your observation with a taxon, not necessarily a specific photo with a taxon, so I’m not sure if this is really a problem, but I guess if other people find this confusing then maybe it is.
An oversight, will fix.
That’s how it’s supposed to work. If “We’re pretty sure it’s” a sparrow but you choose to see only plants, we don’t have a single suggestion to show you that we have much confidence in.
It the top 10 suggestions without filtering out the non seen-nearby options. So basically the same as the production version of the app does now.
Maybe. Is suspect we avoided the verb “Add” b/c we often use that to label buttons that create content like “Add an identification”
Again, maybe, except it’s not really viewing all suggestions, just the top 10, which is an arbitrary cutoff.
When it shows the taxa that are both visually similar and nearby, they’re sorted by a score that combines the vision data and the seen nearby data. When it shows only the visually similar taxa (regardless of whether they are seen nearby), it’s sorted just by the vision score.
There were coordinates but no taxon chosen in the filter. Filtering by a taxon fixes it. Naturally, the broader the taxon, the longer it takes to load. If Life or a large phylum is selected, it will take a long time to think and then sometimes close the image viewer as before. Sometimes it will work and load mallard, honey bee, monarch etc. Trying the same photo multiple times gets different results. One time trying this it closed out and submitted my test photos as an observation. Maybe restrict it to lower level/smaller taxa so people don’t do that accidentally? Or just get annoyed. It works very nicely for smaller groups, but if you select Reset it closes out of the image viewer (which maybe is supposed to happen) but sometimes it crashes and submits the observation. This happened consistently with Bryozoa as the taxon, but not with Rhodophyta on the same photo (and there are way more observations of the latter). Selecting the taxon doesn’t do that though.
Ok, we just released a new beta (build 472, version 1.21.5) that should address the crash you were experiencing, @thomaseverest. Holler if it’s still happening after you upgrade to that build.
OK I think that fixed it. Now it has a mandatory filter for RG search: taxon or place. It defaults to the county of the observation but bringing it up to North America still works. If you add a taxon filter you can remove the place filter, but resetting it goes back to the county filter. Not that that’s problematic (it’s probably helpful), but it’s new.
Currently no, casual obs are not considered in the definition of “nearby,” which will probably lead to a different kind of incorrect identification: people misidentifying cultivated plants as nearby lookalike wild plants (probably some animals too). Our thinking is that that’s a lesser problem than everyone identifying every orange flower on Earth as a California poppy, but it’s definitely still a problem. We may experiment with including captive obs in our definition of nearby, which might address this… or it might seriously mess up our “We’re pretty sure it’s X” recommendations. We’ll see.
