Twitter versus Facebook: Scraping, Faking and Influencing
“There are lies, damned lies, and statistics”
A Company with 100,000 followers is perceived very differently than another company with 5,000 followers. It is very easy to follow a person who already has 50,000 followers than a person who has just 100
I did a small experiment to check the shady world of fan-building. The tools used were AddMeFast (a like-exchange website which allows people to like other people’s social media pages and get their own pages liked in exchange), Python (a scripting language) and Mozilla Firefox (a web-browser).
Like Exchange has become so common that there are Facebook groups which do just this. People charge $5 in Fiverr for increasing likes by 100, 200 or even 500. AddMeFast is one of the many like-exchange websites available today. While it claims to be bot-free, creating a bot is really easy. This article is not about teaching you how to scrape. This is about the inferences that can be drawn observing the bot in action:
- Politicians and music stars in their teens and twenties are among the most frequent users of fake likes and followers both in Twitter and Facebook. It shows that today even if you cannot make your way to stardom, you can fake your way to it.
- It is very easy to fake fans on Twitter. Fooling Facebook is much harder. Of the followers obtained through Twitter, only 10% were detected as fake accounts. Facebook detected and deleted about 90% of the likes. The 10% might actually have been genuine accounts. This might be one of the main reasons why Facebook generates much higher revenues in ads than Twitter. A marketer with a limited budget will spend on Facebook ads because Facebook makes other methods very tough to implement. Twitter can significantly boost its revenues and its market valuation by being more vigilant.
- Finding fake profiles may be like finding needles in a haystack. Facebook’s algorithms are very smart. While Facebook was able to remove about 80–90% of the initial likes, it removed almost 100% the second and third time (through AddmeFast. There are other methods where Facebook is not this good). Facebook algorithms learn which Facebook pages are trying to game the system by identifying some fake profiles and then flagging the pages that those profiles like. This becomes much easier as any new like in the flagged pages could be a potential fake. A profile which likes multiple flagged pages is a fake profile and facebook takes action on it. This is all a conjecture but it best explains the results.
- These platforms could pool information to increase their efficiency in detecting and deleting fake accounts. Creating an API-based service which accepts hashed emails (none of the services would give their user’s emails directly) and then returns a probability of the account being a fake or accepts a hashed email with the probability that the account is a fake might be a surprisingly simple solution to this problem.
The power that social media wields over people today is palpable. Elements are using that power to influence the minds of people. While this cannot be completely stopped, it could at least be made a lot tougher.
“You take the blue pill, the story ends. You believe whatever you want to believe. You take the red pill, create your own bot or reach out to me, and I show you how deep the rabbit hole goes.”
