Add a camera-first species add option to the list survey

[kazlauskis]

From a user:

Please would you add a button that accesses the camera on the List page. The right side of the green “Add Species” button would be perfect with sufficient separation so as not the tap it instead of adding another species. You could have the appearance of the button configurable in the settings, perhaps to the left or right as the user prefers.

This would work well with an image classifier flow in the future.

[kitenetter]

Could this approach also be extended to the standard 'add a single record' option. At the moment, you tap on the plus icon to add a record, at which point you need a species name. OR, you can tap on the camera icon to add a photo, but it isn't very clear what the process is after that, and also it does not provide an option to upload a photo from the gallery.

Would suggest that when you tap the plus icon you are then offered a choice of 'add record details' or 'add photo'.

[kazlauskis]

merge the multiple occurrences from AI

[sacrevert]

I wonder whether any consideration has been given to the quality/quantity trade-off here? By allowing photo-first/AI-naming I think we can expect a surge in incorrect records. This will in turn create considerably more work for verifiers. (I got the impression from Lucy Ridding's eSurveyor field trial that a lot of "false positives" were returned from this sort of methdology). I'm not saying that it shouldn't be done, but it seems like the sort of step-change in data acquistion and quality control that should at least be thought through and discussed before implementing.

[kazlauskis]

@kitenetter currently, the classifier is only used for suggesting search options, which is partly why we haven't sent the AI metadata to the database yet. In line with the approach of the other apps, we will start automatically setting the species value and merging occurrences of the same taxa based on the AI suggestions. Is that OK?

[sacrevert]

@kitenetter @DavidRoy Personally I don't think we should automatically set the species value. Lucy's eSurveyor field experiment found that experts recorded 139 plant species in plots versus 246 for non-expert botanists relying on the app (this is being written up as a paper). The difference (+107 taxa) is almost entirely due to false-positives from the non-experts taking a photo and just accepting the AI recommendation. You can thus see why I am concerned about flooding verifiers.

merging occurrences of the same taxa based on the AI suggestions.

Can you explain this further? I'm not sure what it means.

[kazlauskis]

merging occurrences of the same taxa based on the AI suggestions.

Can you explain this further? I'm not sure what it means.

@sacrevert In some surveys, such as the "Moth Survey," users record a unique list of species with a single abundance count for each occurrence. When the classifier identifies a species from a photo, instead of duplicating an occurrence with the same species (which would have an abundance value of 1), we automatically merge photos of the matching species occurrences and increase the abundance value. This way we can avoid creating multiple occurrences with an abundance value of 1 for the same species. This works differently for surveys that record exact occurrence locations, such as the "Plant Survey" or "Species List Survey.".