DARPA investing $42 million in "memetrackers" | Hacker News Reader

DARPA investing $42 million in "memetrackers"

theatlantic.com

45 pointssethbannon19 comments

Heh, been predicting this for a while. In the short-term a select few companies will know the intricacies of the functions in social groups, how any individual responds given certain inputs, who the social influencers are, and how to spread a message rapidly through any given social network using all this information.

In the long-term, there is the possibility that a group/company could manipulate and in-essence, "program", a human being to think and do whatever they desire. I'm talking here, the ability to dramatically alter an individual's deepest, strongest beliefs, anonymously and automatically. The target individual wouldn't have a clue that they were being slowly and incrementally re-programmed. It could be a global, anonymous propaganda/conditioning system. Brilliant if it were a decentralized system, but I think that would rare.

I think I know how to do this, but I'm definitely not in a position to do anything about it (yet). You'd need huge content distribution channels and access to huge amounts of user data - Facebook could do it. The other option would be to develop a recommendation algorithm system where you encourage companies with existing content platforms to use your algorithms to route recommended content to individuals (now this one is more accessible, and I'm gradually working on it).

I posted a vidcast on the possible implications of the DARPA social media initiative back in September: http://www.youtube.com/watch?v=PrlHVDAjOSU

Media effects research suggests otherwise.

This is a hot sociology research topic and debates about how easily media can influence people have been raging for nearly the past century (well, basically ever since mass media arrived). Empirically it has always been hard to find any media effects at all. Opinions cannot be easily changed by media. Other effects like agenda setting or framing and priming have much stronger empirical support.

This suggests to me that it is extremely hard to control people's opinions and views through media content, especially given the extent to which a single actor can influence media content.

Well I'm making the assumption that our behaviors, beliefs, values, knowledge, personality etc (the algorithm that is "you") are directly influenced by all of our inputs (visual, auditory, sensory etc). Sure there are other inputs during a 24 hour day, but the Internet, and sitting in front of a computer screen are pretty damn dominant.

So if you take that assumption, at face value there should be significant "programming" happening.

Now most of the time, it's yourself that is doing the programming. e.g. you browse content (articles, videos etc) that you actively search for, or actively select out of a feed of possible content.

The idea with recommendation engines is that within the next few years, they will become far more advanced. So now assume that you have algorithms that can serve you content that you absolutely love with 99% accuracy - 99 out of 100 HN stories you're like "OMFG THIS IS AMAZING".

At that point, you will very rarely search for information any more. You'll still do so from time-to-time, but ultimately I think people will sit on a platform and just have content after content delivered to them. And already, the newsfeed nature of the web has started down this path. Facebook, Twitter, Reddit, HN - they're all feeds of information, some with primitive filtering methods. It's in their best interests to serve 100% personalized content as soon as that technology becomes available.

It also doesn't really matter who produces the content. Content will be produced for every single niche. To begin to re-program the individual, you just slowly introduce content that takes the incremental step from their existing belief/value/behaviour toward the desired one.

Out-there example: you have an individual who is a Christian. You simply tell the system that you want them to become an Atheist at X point in the future. The system then subtly, and incrementally introduces catered content designed to facilitate that transition.

Don't get me wrong, this won't be easy to create such a system. I'm just saying it's within reach well before this decade is out. And the concern/fascination is that this could be run globally, and anonymously.

/rant (I love thinking about this stuff) :)

People didn't really search for content when there were a handful or just one TV station or radio station. That's exactly the time when the hypothesis of strong media effects was very popular (also because of the then current fascist or communist regimes and their propaganda) yet researchers were again and again unable to confirm it.

I'm not saying that it's in principle not possible to influence people's opinions but I think it's very hard and we understand next to nothing about how one would go about influencing opinions in a systematic fashion.

> At that point, you will very rarely search for information any more. You'll still do so from time-to-time, but ultimately I think people will sit on a platform and just have content after content delivered to them. And already, the newsfeed nature of the web has started down this path. Facebook, Twitter, Reddit, HN - they're all feeds of information, some with primitive filtering methods.

Partly related, or maybe not, I just wanted to share a piece of advice my grand-father gave to me regarding "media manipulation". To add some context, he was a small-town mayor in an East-European communist country in the 1960s. He really believed in Communism. Anyway, he told me that I should always "read the newspapers endorsed by the Government/official establishment", this way "I'll always know what's about to really happen, or anyway, what the powers-that-be want to get happen".

Strangely enough, it was really good advice. For example I watched this in action during the current Greece debt debacle, with The Economist (an establishment magazine) advising for a haircut on the Greek debt at least one or two months earlier than their CDS spread going through the roof, and at least one year earlier than the haircut actually happening. And other examples are also easy to find.

> So now assume that you have algorithms that can serve you content that you absolutely love with 99% accuracy - 99 out of 100 HN stories you're like "OMFG THIS IS AMAZING".

It's unclear that that is possible. (Perfect face recogniton isn't possible and it's an easier problem.) However, even if it is

> The system then subtly, and incrementally introduces catered content designed to facilitate that transition.

doesn't follow because that would require "perfect content" that also is changing a world view.

I don't get your post at all. What exactly do you mean by 'media' which apparently excludes these other effects? As far as I can tell mass media does all those things. Moreover, one particularly successful mass media outlet is virtually entirely controlled by a single actor.

Media effects studies were done with television, radio, print and web content.

Opinion changes in response to media content would generally be considered strong media effects. Agenda setting, framing and priming are of course also media effects but they are seen as much weaker.

There also still is plenty of pluralism in media content, both internal and external. Worrying developments do exist, sure, but in the grand scheme of things they are not such a big deal. Especially the web can actually increase pluralism.

This doesn't suggest that, as you say, 'opinions cannot be easily changed by media'. Sure, it would not do much good for a personality to appear on the air and basically tell his audience 'you need to think this', but these weaker, less direct methods of influencing opinion are effective and mass media uses them with generally great success. Perhaps the key word is 'easily', but the damage is done regardless.

Agenda setting, framing or priming are not opinion change. It's not really known how and how effectivly those effects can change opinions.

I would also be wary of characterizing mass media as a single entity. That's just doesn't reflect reality.

Rupert Murdoch is the king of News Corp.

Anyway, I see what you're getting at, but I think you're implying a discontinuity between 'change opinion' and 'influence opinion' where one need not exist.

But even Murdoch isn't some kind of tyrant – he can't be, even if he wanted to. There is still a lot of pluralism inside of News Corp. You are, of course, correct: Concentration of media ownership is prevalent and a worrying development.

> In the long-term, there is the possibility that a group/company could manipulate and in-essence, "program", a human being to think and do whatever they desire.

Do you really think this is new? While we haven't arrived yet at Dollhouse-style programming of individuals, programming the masses has been around since the dawn of civilization, and simply increased its efficiency with the advent of mass media and psychology. Now the "compliance professionals" are moving towards a new level of efficacy: algorithmically personalized messaging, rather than scattering seed at random.

Make no mistake about it: all communication is fundamentally mind control. (Don't think of a pink elephant.) We can't stop it; we can only learn to defend ourselves.

Nicholas Christakis has a great TED talk about this called "How social networks predict epidemics" here: http://www.ted.com/talks/nicholas_christakis_how_social_netw...

There's a conference this week in San Francisco that is to cover these topics: http://SentimentSymposium.com

note to DARPA: may I suggest buying Reddit?

Seriously. Reddit has struggled for years under Conde Naste to get adequate funding, and they have already built the community and technology to do exactly this. Then DARPA comes in ready to drop $40+ million, I doubt their project will ever get off the ground, community-wise. Meanwhile Reddit continues to scrounge up a living off advertising.

I'm pretty sure Reddit's userbase would immediately quit as soon as the site was sold to DARPA.

Agreed. I was only half-serious. I do think that what this DARPA project could do, for much less than $40 million, is write software that scrapes a bunch of existing websites and newsfeeds, looking for common patterns that could constitute memes. And that Reddit.com would be one obvious site to have in their list.

I would gladly write such a system for a mere -- (pinky in mouth) -- one million dollars, however.

Reply on news.ycombinator.com