Finding and dealing with fabricated observations on iNaturalist

One of the most unpleasant data curation tasks you can do on iNaturalist, but also one of the most important, is helping find and deal with fabricated observations on iNaturalist. These are instances when an observation is falsified, usually by posting an image found on the internet and constructing an observation (where, when, what) around it. While rare, this practice can be painful time wasters and morale killers for identifiers on iNaturalist, and they can have dangerous data quality and copyright implications.

There may be others, but instances I’ve come across of data fabrication usually come from three sources:

Students under duress - this is probably the most common and occurs when teachers coerce students to create observations as part of their grade and the students think they can get away with it. Generally photos are posted from elsewhere and observations are constructed around them.

Misunderstanding photos as evidence - occasionally new users misunderstand iNaturalist and think that the photo is supposed to be a representative photo of the species they saw rather than evidence of their encounter with an individual organism.

Posting on behalf of others - on iNaturaist, observations are supposed to be your own. Rare instances of posting on behalf of others are tolerated when you’ve obtain permission from the photographer to post their image, when it’s clearly indicated in the observation description that the observation is not your own, and if you have the information about the date, location, and context of the observation and are willing to field any potential questions about the observation from the community. But when people post on behalf of others without permission or a clear understanding of the date, location, and context of the observation, the same copyright and data quality issues can arise.

Spotting data fabrication

While it’s not always possible to spot fabricated observations, the images can often provide clues. Several very ‘high quality’ photos (ie professional wildlife photography) mixed in with other ‘low quality’ (ie blurry iPhone pics) photos is an easy way to spot egregious examples. If you’re using Chrome as your browser, you can ‘right click’ and select ‘Search Google for Image’

This will compare the image against millions of other images on the internet and often reveals that an image does not belong to the observer. Other services like TinEye do the same thing. Keep in mind that this often doesn’t work when the user takes a photo of a photo (from a book etc) that might not be digitized. And also sometimes it really is the observers photograph if they’ve previously posted it elsewhere so be careful.

There are other clues other than photos that observations may be fabricated. Location data is another useful tool. For example, by clicking on the ‘Calendar’ link its easy to see if locations from the same day are scattered randomly over large distances. But in my experience this usually results from user error rather than intentionally fabricating observations.

Flagging the fabrication

If you’ve determined that an observation was fabricated from an photo pulled from the internet, click the (i) button on the bottom of the image on the web observation page to load the photo page.

At the bottom click ‘Flag this photo’ and then click ‘Copyright infringement’

This flag will replace the image with a ‘Copyrighted media removed’ notice on the observation page and cause the observation’s data quality to become ‘casual’

Please give the observer the benefit of the doubt initially. It’s possible you made a mistake and possible the observer was just misunderstanding and not intending to be malicious, or maybe it's in the ‘posting on behalf of others with permission gray area’. I usually leave a comment that links to the Google Reverse Image Search results along the lines of the following:

Suspending Repeat offenders

If someone posts three or more fabricated observations with photos posted from the internet, and is unresponsive when asked to stop, they should be suspended within 24 hours. Please give them a chance to correct their mistake if it is not malicious. If you're a curator you can do this; if not, please bring it to the attention of help@inaturalist.org or the forum. Just before suspension, the user should also be messaged with a message along the lines of the following:

We've suspended your account because we noticed that you've posted numerous observations that appear to be fabricated from photos that are not your own. This can have serious data quality and copyright implications for iNaturalist. Our apologies if this was done in error. If you'd like your account to be reactivated please email help@inaturalist.org for more information. Thanks,

If it's obvious that the student was part of a class and who their teacher is (sometimes this can be deduced from relevant projects) I also usually message the teacher so that they can help address the root of the issue.

Note that suspending a user does not automatically delete or flag their content.

Revised on August 06, 2019 20:08 by tiwane tiwane