Hey, makers and hunters!!
Ever since the beginning of time, humans have co-existed peacefully with bots. In 2017, some twitter bots started to spread false information. Trump, reportedly, has 15 million fake followers that actively contribute to delegitimizing the media by perpetuating his blatantly false claims.
We made this chrome extension to help twitter users be more discerning about determining the legitimacy of the tweets in their newsfeed. Because false realities will only continue prevent the constructive and productive discourse needed to advance our society.
This chrome extension was built using the Botometer API (botometer.iuni.iu.edu), which was built from the models described in this academic research paper (https://arxiv.org/pdf/1703.03107...). I've been in touch with those researchers, who are continuously working on improving their modeling. They're also providing me with great product feedback, such as allowing end users to report bots. So keep an eye out for product improvements!
Hope this helps you. Happy to answer any questions!! Or find me on twitter if I'm unresponsive!!
Report
Can you tell us a bit more about this, looks like a fab idea
It's working, but pretty aggressive... seems like verified accounts should require significantly more evidence. So far Sen Schumer, Mark Knoller, Chad Pergram & Dave Weigel have been flagged as bots with >60% confidence. All verified accounts and all known to me as real people.
@sbmarkb Yes, this is a current limitation of the modeling, e.g. organizational accounts like "@barackobama" tend to score high (you can read the FAQ here: https://botometer.iuni.iu.edu/#!...). The reality is that it's difficult to classify bots. We're working on improving this. Thanks for your feedback!
Report
@andyjiang thanks for that. I did read the FAQ. I’m not questioning your implementation, rather trying to understand. For instance, Mark Knoller & Dave Weigel are reporters; they write their tweets - they just happen to be writing about trending topics with high frequency. As a PM, I’m curious as to the decision not to implicitly trust a verified account as “not a bot” when classifying the content. I can’t think of a verified bot, but if they existed it would seem there would be significant false positives (negative value) in the detection of a few accounts. I tend to follow reporters, politicians, musicians... 2/3 of those being detected as bots renders the tool unusable for me.
@sbmarkb Hey no problem. This is great feedback, so thank you for sharing. Understand that it's unusable, hopefully we figure out the right way to improve the model :)
Report
@andyjiang what a great idea you have going, though, and rest assured I have the extension still installed. Will definitely keep trying it out as you work out the kinks.
Botson
Botson
Postmates
Botson
Botson
Botson
Anvil
Botson
Honor Care Network
SentiOne