Hi everyone, I often ID audio on iNat. One user regularly posts observations of plants, lichens etc. that include both photos and audio. The photos are mostly decent, but the audio is invariably short (often 1s) and has no obvious relation to the taxon (i.e., this is not a “reeds rustling in the wind” situation, the audio seems to be entirely random). Many of the observations reach Research Grade due to the photos. I’m not really sure what to do with such observations as an IDer. I have asked the user to stop adding unrelated audio, but they continue. Should I be clicking ‘No’ to ‘Evidence related to a single subject’? The audio doesn’t relate to any other subject as it’s just noise, so this choice feels inappropriate. Doing nothing is also not ideal, as then the audio becomes associated with that taxon. (Maybe not a fundamental problem for taxa that aren’t regularly IDed by sound, but it still results in a bunch of rubbish being added to iNat.) Any advice about how to deal with this situation?
Have you asked them why they’re including the sounds? If so, have they responded at all?
No, I admittedly haven’t asked - the audio seems so obviously pointless that I just asked them to stop. (To which they didn’t respond.) I can ask them though.
It would be interesting/helpful to know why they’re doing this.
Are they a part of a project?
Two years ago, I volunteered with a cicada soundmapping research study that required us to make recordings at specific locations over a particular period. Many of these ended up having no evidence of cicadas and ended up labeled as other organisms by identifiers.
If they have a large number of observations that include audio of the same length, it may be intentional to satisfy a requirement.
It could be that they were trying to upload video, or didn’t realise they were uploading a video, and it presented as audio only. I have a file on a multi-photo observation that I can’t seem to delete, and that is the reason it is appears as an empty audio file in the carousel.
Given that they’re all observations of pressed or otherwise collected plants/lichens mapped to the date and location of collection with a 2-second long clip of nothing, it appears that the audio recording is being used to document the date and location of collection, and the collected plant images are then added. Of course that’s all just speculation until they respond with their own explanation though. (Sorry if anything I’ve said is considered enough to “out” the observations in question, but I was already able to find them easily using what was given in the OP already, so…)
I was wondering the same thing. On iPhone you can shoot in “live” mode, and it creates a video with sound.
Thank you all for the theories as to why. The one from @paul_dennehy sounds the most plausible to me in this context. I have asked the user why. Any thoughts on what to do if they don’t respond in, say, a week? They’re a fairly active user.
If they’re using audio files to document the location, the user should be able to delete the audio file after uploading while maintaining the correct location. Or at least, that has worked for me with photos of collected bryophytes – upload a photo of the site, then upload the photos of the collected specimen, then delete the site photo. (Except that more often I leave the site photo in the observation as a record of habitat, substrate, etc., then change the default photo to one showing just the bryophyte, which leaves the location data at the correct location. But everyone has their own workflow.)
My understanding is that Computer Vision does not do anything with audio observations at this point, so at least they’re not messing with the CV model.
As for users adding “rubbish” to iNat – it’s definitely annoying (speaking from experience, e.g., IDing “Unknowns” contained in a high school class project usually turns up quite a few seemingly random photos!). When I get annoyed by those seemingly random photos, i just say to myself, “Well, it’s better than Facebook or TikTok.” Low bar, I know, but still…
Ok, so the vibe I’m getting from this thread is don’t sweat it and just leave those observations at RG, even thought some media don’t relate to the observed organism? I’m ok with that if it’s the consensus. For plants I guess it’s fairy unlikely that unrelated audio on RG obs will lead to issues down the line. For animal observations, this kind of thing might become more of a problem if iNat starts training the CV model on audio (I’m not really sure why they don’t already?)
Why should audio be treated differently from photos? They are both evidence and if they are unrelated, why not use a DQA?
Which question in the DQA would you use @zoology123? The only ones that seem potentially appropriate are ‘Evidence related to a single subject’ and ‘Evidence accurately depicts organism or scene’.
Definitely Single Subject if you’re going that route. That’s meant for when someone uploads media that don’t all consistently depict the same organism. It’s most commonly used when someone uploads a bunch of pictures of different stuff they saw as one single observation, but it arguably applies to a photo and a sound uploaded together that don’t depict the same organism.
But I would only use the DQA for ‘not a single subject’ if you can hear something other than the subject of the photos - a cricket chirping vs 3 pictures of a tree.