Finding and dealing with falsified observations

At the risk of distracting from an very awesome CNC - I did notice a few instances of students involved in CNC fabricating observations (I know, bummer topic) which prompted me to write the following wiki-page for finding and dealing with this content. I also linked to it in the curator guide. Curators, please edit/expand as needed: https://www.inaturalist.org/pages/fabricated_data
And thanks for all the awesome IDing and data curating work - you know who you are!

7 Likes

Thanks @loarie,

A while ago I flagged a bunch of images as being scans or photos from books, but I didn’t know how to deal with it. Your post and wiki was a great resource.

(Now, if only there was an easy way to deal with the ‘students under duress’ issue in general…)

@loarie - do you have any recommendations for dealing with the situation when a single user has a large number (beyond the suspension). I’ve just noted a user who has added over 500 records for the CNC, every one that I have checked so far is a copyright violation.

That being said, I’m not hugely enthusiastic about checking and flagging 500+ records.

Just to follow up as I go further down their records, some appear to be their own records, which in a sense is even worse, because its a mix of legitimate and stolen.

1 Like

Thanks for doing! Very helpful indeed. It really didn’t sink in before that people may do this, and as I think back on the IDs I’ve done over the past 4 or 5 days, there certainly have been obs where I’ve thought “wow! what an awesome pic!” Hopefully most of them really were from the poster…

So all that said, I realize I don’t know how to run a view of all the observations (other than my own) that I’ve done anything to (ID, comment…). Does it require a url-level command, or am I having a senior moment over something that is very obvious!?

Theoretically this should give you a list of all records you have interacted with (obviously switch out my user name for yours)

Please note I am unsure if there are any delays in getting things accounted for there due to the CNC crunch

https://www.inaturalist.org/observations?place_id=any&reviewed=true&viewer_id=cmcheatle

UPDATE - there clearly are some delays as the most recent things I have done do not appear. I am unsure how long it will take for them to show in the results.

1 Like

Seems to work fine - thanks much. And yes, I hear the caveat - even “Unknown” observations I pull up get updated between running the query and getting to that observation. And I would think that the only consequence of finding the same ones you did would be double-alerting the staff, right?

I’ve had major issues trying to look and ID observations the last few day but if I see something I’ll flag it

I just got bogged down in the same user’s observations. Several don’t turn up any results on Google Image Search, but are of species that would clearly not be present in that location (Common snowdrops in the tropics, for example). I initially wasted time checking the location to see if it was somewhere like Fraser’s Hill, where conditions might have been right for the British to have introduced them as a reminder of home back in the day, before coming to this thread to find out what to do in cases like this (and discovering that not only “cases like this”, but “this specific case” are highlighted here). What is the recommended action to deal with this - I’m guessing that this user isn’t going to be reading comments attached to these observations? I decided that a personal message probably wasn’t appropriate in this type of case.

My personal view (others can chime in) is that if you can’t prove that a photo is taken, you can’t remove it on under the assumption it is taken. That’s why I left those ones in place.

You can vote in the DQA that the location and or date are not accurate because that is your assessment of the accuracy of the data, and leave a comment to that effect. One challenge is often the ID is not accurate either as they just run the computer vision and pick a result.

In the category of ‘sometimes you just can’t win’, I just spent an hour removing violations from a user.

Who was it ? The same user highlighted in Scott’s tutorial who just created a new account and started again, in fact even with some of the exact same photos…

What do I do when I find two different people using the same photo?

1 Like

If it is noted and credited as being from another user, that is considered acceptable.

If I understand correctly, this user has submitted some legitimate and some illegitimate observations. Analyzing which are which would take a considerable amount of identifier or curator time that could more productively be spent in identifying observations from non-cheaters. I would suggest that everything from this user be discarded. Can we do that?

There is no tool to do it. My personal view is that curators or anyone who want to flag something as a violation need to be able to prove it. Suspicion is not enough.

At times if i cant prove it but suspect it may be a stolen photo, I will put in a comment noting the user history of doing so and that any id should be done with caution. But I won’t remove it.

2 Likes

Well both observations “look” legit - one person has a number of observations, and the other person, only has the one that matches the other’s photo. They are exactly the same photo. I would be tempted to flag the person that only has the one observation.

I have a related question that happens to have just come up so I thought I’d ask here since this post is still pretty recent:

What do people do when they find what looks like users cheating (for lack of a better word) on these assignments by posting photos that (I assume) are not actually what they claim to be? There is a university class project that has been generating a lot of observations - it seems they need 10+ observations of a variety of life (fungi, mammals, arthropods, plants, etc.), which is cool and I’m glad to help of course. We came across a photo associated with this project that appears to be an Australian (or otherwise non-native to North America) orb weaver spider. Easy to spot since there are only a few species of this very-recognizable genus in the area where it was listed and this is clearly not one of them. Looking through the user’s other observations I found several other suspicious ones. The questionable observations all have “Screenshot” as the source in the EXIF data, whereas the user’s other observations of common local things were all taken with the same model of iPhone. So kinda seems like they ran out of time on their bio class project and just found some photos to pad their numbers. Either that or they got really lucky to find several non-native species on the same day.

My question is: Is it inappropriate to report these to the project owner or maybe even track down the course instructor? I’ve never taught university courses or dealt with this sort of thing before, and again it’s just a suspicion. I couldn’t find the source images with reverse image search. A couple people mentioned it in one of the observations (something like “did you really take this photo in [state]? that would be an amazing find!”), and the user deleted that observation the next day, so that kind of adds to my suspicion. What would you do here? Is it none of my business or should I tell someone?

Hi @jgw_atx, it’s not everyone’s cup of tea to deal with these, but it’s important that false observations be marked and hidden. There is a tutorial and post about this here: https://forum.inaturalist.org/t/finding-and-dealing-with-fabricated-observations/2758

2 Likes

Thanks. Since I couldn’t find the source of the images I can’t flag them for copyright infringement so I may just try to relay this info back to the project admin. Appreciate the reply.

If nothing else, you can always add a comment to the record that questions the source of the photo (if you can see that it was likely lifted from somewhere), which might be the quickest way to deal with it if you don’t want to chase down a project administrator. At least subsequent visitors would then be forewarned the record might have problems.

4 Likes