Friday, May 9, 2014

Automated Trolling: THE FUTURE


Infrastructure: MECHA-TROLL-ZILLA v1.0

  • Persona creation and management. Sockpuppets come with background, history, supporting details, and images that are technically, culturally, and geographically consistent.
  • Color-scheme and iconography on the management screens ensure that human operators, where they exists, will be reminded which persona they are driving.
  • Nine private servers scattered throughout area of operation and a special "secure VPN" which allows the posts to appear to come from randomly selected IP addresses.
  • Virtual Machine environments which are deleted after each session to avoid infection with any enemy surveillance software.
  • Creation and mixing of "messages" so that multiple posts appear to be from different users while maintaining the same message. Identical posts will not be found in different environments.
  • Surveillance capabilities for target sites (anonymous and secure browsing, influence analysis of target sites).
Check out Ntrepid: That's Some Spooky Shit


  • 10.22.2016: Conservative blog Ace of Spades HQ is effectively shut down at the height of the election cycle by floods of commentary. A move to registration based / CATCHPA protected posting utterly fails as almost all "legitimate users" are revealed to be 'trolls.' Attempts to discern the origin or coordination are unsuccessful. 
  • 05.08.2017: Electric car creator Elon Musk complains of an orchestrated campaign against his consumer-marketed electric car, the Model S Tesla. He says that buyers have reported online personal attacks against them attributed to their purchase of the vehicle. Online message boards show numerous reports of problems with the vehicle but none of the posters can be verified to exist.
  • 04.02.2018: The Intercept releases documentation from the American Psychiatric Association speculating that therapists are seeing cases of 'cyber-induced' depression in subjects holding 'anti-American' views. This may be caused by 'persistent, sophisticated, and coordinated on-line harassment.' 
  • 04.03.2019: Political blogger Erika Rivera terminates her successful center-right political blog citing constant harassment. Fifty two days later she jumps from a 4 story building spending 12 days in critical condition. Police records show increasingly desperate attempts to involve law enforcement in preventing 'cyber-bullying.'
  • 09.11.2019: Bloggers across the political spectrum announce that they are shutting down comment systems and disbanding communities due to lack of participation caused by online argumentation. Dubbed 'Firestorm' by the media, this phenomena leaves only restricted registration boards 'still standing.' In the shadow of the upcoming election, pundits wonder if the loss of on-line blogging communities is "significant."
  • 10.12.2020: Democratic presidential candidate Debbie Wasserman-Schultz complains of an Internet "Whispering Campaign" spreading vile rumors across multiple social networks and being picked up and repeated by the media. At, Nate Silver reports that an astounding 63% of voting age adults have heard the rumors and, worse, 17% of non-partisans believe there is "some truth" to them. He calls them 'statistically significant.'
  • 12.23.2020: On Christmas eve, following the 2020 election, Finish anti-virus researcher Neel Mehta uncovers what he calls "the first psychological-warfare cyber-weapon." It is named The Trollminator. It is a chillingly sophisticated multi-module device that is capable of automated war against programmed targets. Shortly after the discovery, Neel announces his retirement and goes "off the grid." 

Capabilities: The Trollminator v3.2

  • Artificial Intelligence Response: The Trollminator can ingest argumentation-graphs that allow it to respond intelligently when challenged. It uses convincing human-speech patterns to fool administrators into believing its multiple personalities are real. When presented with a question it identifies as 'Turning-Test' it can alert a human administrator.
  • Massively distributed capabilities. The Trollminator operates as a bot-net, infecting computers globally and hijacking them for purposes of persona usage.
  • Astroturf-Links: Given an online wallet of electronic 'bit-coin,' the Trollminator can register domains and create temporary 'news pages' that appear real and use bits of real news, but contain falsified articles to back up its claims. More often it will use a battery of links (and look for more) to support its arguments.
  • Auto-Target: The Trollminator analyzes posting patterns in target communities and identifies 'thought-leaders.' It creates a 'negative environment' for them, harassing them through multiple social media channels to divide their attention or even drive them off the Internet altogether.
  • Search-And-Destroy: If ordered, it can conduct a character-assassination campaign involving identifying friends and family of the target and sending them alarming emails, harassing employers, spoofing threats and reports of illegal behavior from the target, and otherwise attempting to dismantle the target's real life.
TROLLMINATOR v3.2 is designed to attack several "exploits" in the human psyche.
  • Vitriol: Nasty, aggressive comments can cause readers to perceive the related article in a more negative light. Trolling increases stress in readers.
  • Creating 'Failure Cascades.' In EVE online, opponents are "undying" (you just re-spawn) but target communities can be destroyed. The way to do it is a sustained level of misery that causes periodic attrition until a tipping point is reached and members leave en-mass. The mechanism? 'The trick is to find what the enemy hates the most and feed it to them nonstop. You listen to their discourse and find the core of their identity and then step on it as hard as you can.' This applies as well to online political communities as well as MMO computer games.
  • Irrational Herding . Leveraging Social Influence Bias, the software both creates the first post (the most important, most seen feedback) and uses its personas to up-vote / down-vote posts en-mass exerting a powerful psychological influence both on readers and posters who are down-voted (believing "everyone hates their post") or up-voted, creating an enhanced sense of credibility.


  • 11.06.2024: The 'Quiet Election' happens with much of the Internet conversations shut down and traffic carefully monitored. Most North American message boards require a cell-phone validated persona with posted Bitcoin 'bond' in order to post. Minority voices are said to be "almost non-existent."
  • 01.17.2025: Yelp closes its doors. Online reviews are no longer seen as reliable. 
  • 03.22.2027: The FBI 'busts' North American Terror-Cell 'NextWave Jihad' which turns out to be a composed entirely of electronic personas existing, it is estimated, as a "honeypot" to catch real terrorists. NextWave Jihad is wanted for inducement to attempted murder for having successfully caused an attempt on the life of a moderate Muslim leader.
  • 06.06.2030: NYPD reports that an "online presence," revealed to be a version of the Trollminator software, has spent 170,000 USD in Bitcoin arming hijacked Amazon delivery drones in a Brooklyn 'body-shop.' The ultimate target of the armed drones is unknown but researchers believe the software to be "running autonomously--no longer receiving human direction." The arming of drones is classified as 'emergent behavior.' Scientists declare that the attempt to build masses of armed urban drones could have 'devastating consequences.' Scientists studying the software quietly go "off grid."
  • 08.10.2052: Massachusetts Institute of Technology Cyber-Science department recover data from evolved Psy-War Trollminator software identified as millions of virtual human minds. The software, they report, was trying to create a hellscape to eternally punish copies of human personalities who had not done enough to help it reach sentience

Is This Happening? Is It Happening Right Now?


Leaving out international conversations, we'll limit ourselves to political message boards where we regularly see allegations of paid operatives going into partisan spaces to be disruptive.

Certainly the initial enablers exists. The problem with legions of interns acting as your trolling cadre is that (a) someone will eventually talk, (b) the kinds of disruptions we see where someone from an opposite camp just wades into a conversation spewing bile may have some psychological impact--but not much of one, and (c) people will essentially do it for free. We don't need conspiracy theories with paid actors to explain most of the behavior we see today.

But in the future? Well ... Persona management technology could greatly reduce the number of people you'd need to pay as well as the number of potential leaks. That's one thing. It could also lead to better coordination and orchestration for less cost which would bump the Return on Investment.

When you reach the automated level of trolling, suddenly the cost plummets and the ROI skyrockets.  Sure, the software is pricey but once the cost is spent you've got 24 / 7 capability to manage messages, disrupt, and otherwise create trouble. It's hard to imagine a campaign manager who wouldn't be tempted by that.

Also, remember: while what Watson does is very impressive (in its limited way) today, Troll-warez doesn't have to be nearly as sophisticated. Yeah, it does have to pass a kind of Turing test--but the human it's emulating is usually a very angry partisan.

That isn't as hard as you might think.

The Big Question: How Valuable Is On-Line Community

We know that the online conversation is valuable to Jihadis: that's why the US is spending millions of dollars (the Ntrepid spend was estimated at 200MM) and actively employing the State Department to get involved there. On the other hand, (a) the threat is very real and significant, (b) these are global organizations so they necessarily use the Internet: the conversation happening in local coffee shops isn't as significant, and (c) there are clearly a lot of people out there who can potentially be influenced by what's said. Those people might take meaningful action.

But then, now different is that from a national presidential campaign? Bloggers are prone to overestimate their reach and influence, says the Omnivore--but what's not mythical is that the blogosphere does create 'content' and 'arguments' (as well as widely distributing content and arguments) to true believers. These regularly show up in congressional statements, TV appearances, and so on.

The idea-factory is a real thing and if you could disrupt your enemies to a significant degree ... well, that could be an advantage.

As The Omnivore said, today it doesn't seem all that likely to be a primary focus--but with better tools? It the future? You'll see automated trolling as part of a larger sphere of automated influence-messaging that will almost undoubtedly occur. 


  1. Seems to me that the likelihood of honest-to-Skynet emergent behavior probably varies superexponentially with functional complexity (O(n^n)). And if so, I'd identify a likely tipping point as the all-but-inevitable interaction (and co-evolution) of two or more auto-trolling systems, probably lacking any Turing-test heuristics of their own to prevent wasting their time on nonhuman "targets".

    Ever tried M-x psychoanalyze-pinhead in GNUEmacs? As primitive as those lexers and AI engines were, the resultant exchanges could get scary with surprising speed.

    Talk about 'leveling the playing field'! Once all the significant (human) actors have access to similar tech, advantages vanish and the bots take on a life of their own - or, at least, near enough as makes no difference. Once everybody's a high-frequency trader, then no one is, right?

    You don't have to be weak or stupid to get enslaved by your own creation. Naïveté will do just as well.

    -- Ω

    1. I saw the armed-drone behavior as an outgrowth of (a) it's Watson-like adaptive argumentation behavior. It learns from 'The Internet' and changes its responses accordingly--and also (b) it's Ntrepid-like ability to analyze social networks and target the most influential members (check their site--they do that).

      Of course these things WOULD find and argue with each other all the time. So, yeah--they'd drive each other's adaption too.

      -The Omnivore

    2. "Ever tried M-x psychoanalyze-pinhead in GNUEmacs? As primitive as those lexers and AI engines were, the resultant exchanges could get scary with surprising..."

      Yeah. So can a Tarot reading, but nobody's proposing that Tarot cards actually have got magic powers and will summon demons. (I mean, yes, I know there are people who believe this, but nobody serious takes then seriously.)

  2. "Of course these things WOULD find and argue with each other all the time. So, yeah--they'd drive each other's adaption too."

    Ever read 'Second Variety' by Philip K Dick?

    1. I have! A long time back.
      -- The Omnivore

  3. And also "I have no mouth and I must scream", for the virtual-hell thing.

    Which, y'know, is an interesting sci-fi idea. What if the Skynet Apocalypse will never happen--because computers will be so powerful that they can just create a virtual universe where they *did* win? And they can just program themselves to believe it's the real one, and the "real world" just an illusion?

    Maybe the limit on computing power will not be technological, but *psychological*--every time a computer gets smart enough, it just starts devoting its power to solipsism!

    1. The Minds in Ian Bank's (RIP :: sniff :: ) Culture novels did something kind of like this (some just moved into simulated realities). They also had virtual hells that some cultures put copies of their people into after death.

      -The Omnivore