We need an API to validate our bot/troll detection techniques.
What I would like is the ability to use an API to alert Twitter to the presence of malicious actors interfering in elections. Ideally, it would be process whereby we crowd source and gamify the detection pieces, with Twitter's decision to suspend (or not) as the ultimate validation determining who is best at the game.
Recently, I sent 500 Canadian election Twitter IDs our model saw as bots/automated to a third party scholar: Kurt Wirth <kurtawirth@gmail.com>. I asked him if I could share his response to what I sent, he said yes. His response is below:
---------- Forwarded message ---------
From: Kurt Wirth <kurtawirth@gmail.com>
Date: Fri, Sep 27, 2019 at 4:18 PM
Subject: Re: Bots
To: Shulman, Stu <stu@texifter.com>
Hi Stu!
Here are the results for the first set. In this set, 28.7% of the accounts are estimated to be completely automated (as shown in cap.universal) and 66.4% are estimated to be at least partially automated (as shown in scores.universal). This number is very high. In other research, I've found approximately 6-6.5% of all accounts are bots and in political discussions the number drops to around 3%. I will get your second set started now.
