Came across one the other day, just by mentioning in a post "blue plaque", got instant like and retweet.
This started me thinking of a little game - what post with the limitations of 140 characters could you post that would trigger the attention of the most bots for automatic retweets/likes?
It's on my to do list, when bored, but I'd imagine others have had comparable thoughts and may of already had a go.
I was trying to build an online presence as an indie game dev for a while, #gamedev and #indiegames generated a lot of bot retweets. Bot retweets tend to cascade, too - other bots that may have missed the original tweet pick up the retweets. Finding the right hashtags can be a pretty good way to get (human) eyes on your message, due to that exponential growth factor.
There's a reason for that. Twitter offers an API to get a massive stream of tweets. They used to provide it in various sizes, leading up to what was called the "firehose". Now they only have a streaming API with filtering (but gets a very small, but constant random sample of tweets), and the "firehose", which is all of them -- and you have to pay for it through a third party partnered with Twitter. When you're doing large-scale datamining, the firehose can be invaluable, albeit expensive (Depends, but is anywhere between $300-$6000/month). But bot makers don't always have that kind of scratch, so they will just consume the public API, which is generally good enough.
My bot is listening to this and it is really awesome! I was surprised to learn how many people on Twitter daily post 'i am bored' up to 300,000 per day! With that amount if data it is readily possible to build robust ML systems.
Mention Beiber and you'll get a swarm of likes, Rts, and responses. It used to be a personal game for me to get a friend to mention him and watch the swarm.
It used to kill the usefulness of your notifications for at least a day. ;)
Better title: "Up to 48 Twitter accounts may be human"
Snark aside, there is no way around this. Followers in digital public spaces are representative of social status, the demand for appearing popular will never go away and so neither will the bots.
This isn't necessarily true. If world governments could use their records to create an OAuth-like service, you could uniquely verify citizenship without giving away personally identifying information
Shall I go ahead and start the over/under pool on how long it would take before it was hacked? I think it's a compelling idea, but it would probably be the most-interesting target in the world.
Depends on how it's implemented. As always, the devil's in the details. If it were properly sandboxed, it seems it shouldn't require anything more than the government's unique ID eg a Social Security Number as a backing for tokens granted. In that regard it wouldn't be any more juicy than any other common store of SSNs, of which there are sadly quite a few.
You would presumably be able to use this information to de-anonymize previously created tokens, but that would require you to gain access to the token as used, which would also require a breach of Twitter or whoever else. It's not unthinkable, but if Twitter is breached you might have enough info to de-anonymize its users anyway.
While automatic control of computerized interfaces should surprise no one, “bot”-like behavior is not limited to computers and isn’t new. One should always be skeptical even when something comes from humans: for instance, a “customer review” could be a paid reviewer, or someone talking on the nightly news could be essentially controlled by the puppet strings of their organization and not really voicing their own position. Ideally no one ever puts too much faith in what they hear from a handful of people because that is not statistically significant (there are literally hundreds of millions of people in the U.S. alone, billions in the world; sample sizes must be gigantic).
Where “millions” of accounts become dangerous is when trying to determine if even a “statistically significant sample size” is legitimate. I seriously doubt that most organizations ever come close to finding a large enough sample in the first place before reporting a finding. And yet, if they primarily look at online postings instead of actually talking to people, it’s likely that even a large sample is totally messed up. That should concern everyone.
And one of them is mine. It recommends an event once you tweet to her. It is actually a service that many people appreciate. So in summary, bot != bad.
To compare, I looked up the estimates[1][2] on Facebook and if we use 170 million out of 1.4 billion, that's ~11.7%. That is in the ballpark of Twitter's 15% (48m/319m).
I guess if you make a popular service that doesn't require a credit-card payment or phone SMS text as a test for human verification, you should expect >10% of fake accounts even with state-of-the-art heuristics detection.
So wouldn't it be relatively simple to block AWS/Azure/etc ip blocks AND duplicate accounts from single addresses? Wouldn't that be a large drop in bots right there?
Corollary: I once played a large MMO that intentionally inflated user numbers right before they were acquired via content updates that encouraged the creation of multiple bots per actual player. I'm sure at some level executives are aware that bots are good for business and allow them to persist.
This is a result of what I call "audience economy". People want an audience and are willing to pay for it. Audience likes to see a prior audience. Bots enable prior audiences. Platforms like fiverr enable bots.
Get 5000 twitter followers for $5
Those 5000 fake followers make you look good to people that come across you, and they flow.
It feels very discouraging to start out in this space once a given platform reaches a certain level of maturity. For instance, Insta now seems littered with these local service/small-business accounts spamming the most optimised hashtags with inspirational quotes on stock photographs; most engagement seems to come from bots and scripts being used by these same accounts. (Maybe I'm playing the game wrong.)
Does buying 5k followers really help in any way? Won't the more sophisticated users (ideally those whom you'd be targeting as followers) see through this based on your engagement rates, or does having a certain number of followers hold some weight with the algos which leads to your stuff attracting more engagement, followers, etc. in some kind of virtuous cycle?
Twitter never sounded like a great idea to me. Years went by and I finally made an effort in the last few months to get on it so I could follow political tweets. I found that it sucked in all the ways I thought it would suck. When you find a good tweet, you get to enjoy reading it for a few seconds, but it leaves you empty and searching for another one.
I think there's a general feeling that Twitter has not much growth potential and it's slowly becoming obsolete.
If it's filling up with bots, that strikes me as a sign of decay.
Unfortunately no. When a social platform becomes "big" as in, reaches early majority, spam finds a way in. No algorithm can stop it. Quantity kills quality. The "sophisticated" early adopters lose it.
Look around. People will do whatever it takes to get to "front page", "top stories", and such.
Medium obsesses about quality, but their top stories, fueled by recommendations from "the common man", are full of junk. Same with LinkedIn. Twitter. Facebook.
Look at Imgur, it has become a series of memes, reused content from (Twitter, FB, etc), or repetitive "dumps". original content is automatically downvoted
I'm thinking that this is not as `black&white` as said.
There are accounts which actually are used by people/companies, they post authentic content but do some automated actions.
E.g. a company account which tweets authentic content for their field but on the other hand has automated retweets on certain hashtags in order to gain their authors as followers or to come up as related to other accounts in same field.
Also to mention that there are services such as IFTTT which actually can help you to automate such actions and make bot-like behavior.
I guess the same would apply to Instagram, Pinterest, Reddit, etc.
There's a method to that madness. Whether he is actually aware of it or not (I lean more towards the former), that stream-of-consciousness styled rhetoric is incredibly magnetizing to a large percentage of people. In my research on AI-driven political engagement, I found that while intensively complex generated argumentation can engage relatively sharper audiences without arousing suspicion, it is more computationally expensive, and those people do not have as much flex in opinion. The use of simpler, "like output of a Markov chain text generator" styled tools help separate the wheat from the chaff in terms of pliability.
This won't suprised anyone that uses twitter. In fact I rely on 2-3 bots to retweet my tweets and so get more coverage. They have a place, it's not all bad.
There's really no legitimate reason why twitter allows follows via their API[1]. I'm not sure if they see some benefit from allowing spam-follow bots to run wild or what they could possibly be thinking.
There's also no reason to believe that botting is done via API in most cases. I would think it's easier to detect + ban that way. I don't know how it's done in the majority of cases, but I know it's not particularly difficult to write a bot that drives an actual browser and emulates human mouse behavior.
This started me thinking of a little game - what post with the limitations of 140 characters could you post that would trigger the attention of the most bots for automatic retweets/likes?
It's on my to do list, when bored, but I'd imagine others have had comparable thoughts and may of already had a go.
Anybody got any other examples?