Fame for sale: efficient detection of fake Twitter followers | Hacker News Reader

Fame for sale: efficient detection of fake Twitter followers

arxiv.org

35 pointspatomolina12 comments

I have no desire to buy "fame", but I do wish Twitter wouldn't correlate the number of people I'm allowed to follow to the number of people that follow me. I frequently hit their cap, and am forced to scan my following list for accounts that have stopped tweeting. This seems like a poor way of handling that scenario - you penalize the person that is still using the service that actively follows the constantly evolving network of users.

I would really like to be to run this sort of analysis (or use tools/services that perform it for me), regardless of how efficient the current popular methods are. Is anyone familiar with the tools or services available to do so? Thanks in advance!

You can try twitteraudit.com (I made it a few years ago). It doesn't do anything fancy at all, but it works pretty well with about a 10% margin of error. It only looks at a sample of 5000 followers though.

Why fake followers? Social proof of course, one of the core tenants of getting a potential customer to trust you.(1)

Faking followers to create the impression of social proof / trustworthiness should, of course, be regulated since it's flat out trickery and falsely persuasive.

I don't envy the FTC's job on regulating commerce with issues like fake followers and the chaos ahead in regulating fake reviews, thousands of undisclosed endorsements via podcasts / you tubers / etc

(1) https://en.m.wikipedia.org/wiki/Social_proof

I don't find a high follower count suggestive of anything. I think that your threshold for regulatory intervention needs to be much higher. Vanity metrics are not a threat to the marketplace.

Sure, maybe consciously you aren't affected, but undoubtedly you see a difference between 0 followers and 10k followers (even if I bought all of them).

Absolutely. Once the govt takes an interest in keeping follower counts "accurate" it's going to meddle with all sorts of other metrics in your app too.

I'm curious if you use ad blocking? Why or why not?

Twitter causes the problem by limiting the number of people you can follow. For example, people who want to grow their user base follow a lot of people so they can getting follows back. You might follow 500 people so you can get 50 of them to follow you. But once you follow 2,000 people you reach Twitter's limit and from there the number you can follow depends on how many people follow you. This is an incentive to buy fake followers, so you can continue to follow the people you actually want.

Introduce the concept of verified and unverified number of followers. If you're willing to lose "verified number of followers" status, allow someone to pay and fill out any number they want in the followers field. If they want a verification of the number of followers (so that marketing companies feel comfortable paying celebs to sell products) have them pay a fee. Its a win both ways.

To call it 'dangerous' is a bit of a stretch..maybe annoying. The easiest detection system is that the tweets of an account with a lot of fake followers will have very little, if any, interaction . I'm kinda irked stuff like this makes to arxiv but you still need the 'two referrals' system to upload stuff. I can't image the worst math paper is much worse than this.

I agree that the easiest detection system is to look at engagement of tweets...I have a few thousand followers and I average about 2 to 5 favorites/retweets...so if I see someone with 50,000 followers and less than that average, I get a little suspicious. But most people don't think to look at this...and news outlets are quick to jump on what seems like a valid tweet from a valid account. The reason why we don't see much of this happening is because most of the people who buy followers probably aren't hoping to do anything particularly malicious or disruptive.

Reply on news.ycombinator.com