Using Benford's Law to weed out bots or 'unnatural' garbage posts/posters

The CredSperiment Research
1

Folks I am new here but not new to sourceCRED. I posted this in another thread but figured to bring it up a bit because it may be an important tool for how to identify and eliminate bots and spam from these initiatives.

I recently saw episode 4 of connected on Netflix (no this isn’t a plug for Netflix) Just providing a link in case anyone is interested that already has Netflix.

https://www.netflix.com/title/80097225

In this episode I learned of Benford’s Law https://en.wikipedia.org/wiki/Benford’s_law and its use in detecting bot activities on social platforms, fraud in taxes, skew in dataset data generally, photographs, videos, you name it. I am still reasearching this topic as I find stuff like this personally interesting.

I wonder if anyone in the sourceCRED initiative is using Benford’s law to identify un-natural actors in the sourceCRED systems and depreciate their CRED etc.

BTW: I am an avid fan of sourceCRED and watch with great interest the activities on Maker forums as a participant in their sourceCRED trial and wait to see where in my on-line life sourceCRED will pop up next.

1 Like
2

Very interesting! Afaik, we haven’t yet got into advanced tools for detecting gaming. So far we’ve been focused on communities where it’s relatively easy to see. But I have already started on some data viz in python that can be used to visually detect anomalies, which is apparently a common fraud detection approach. Will have to look more into Benford’s law.