Fake social network profiles to spread propoganda, impact on the Social Media Monitoring companies?

[monica nixon]

Without a doubt, many of you have heard about CENTCOM's Ntrepid contract for gov intelligence analysts to be able to create sock puppet(FAKE) identities on social media sites to spread propaganda and influence opinion on matters critical to the US administration.

What interests me most, is how long it will be before private firms with savvy CI/Counterintel folks get their hands on this type of SW, and how the TNS Cymphonys, Radian 6s, Visible Technologies, Biz360s, Crimson Hexagons of the world (ie the social media monitoring/brand & listening platform firms) are going to handle this overt manipulation and corruption of their inputs that would then skew their sentiment analysis, volume and tone metrics (many of which are not particularly automated in the first place). Knowing a few things about the technical intricacies and limitations of many of these platforms, I don't believe they are equipped to handle this situation if it ensued broadly which could pose a real threat to the industry. Even if these type of SM firms had broader, automated demographic profiling abilities start making their way into the platforms would it would be sophisticated enough to rat this out?

Makes one wonder about the future of Social Media Monitoring indeed....
Thoughts?

Regards,
Monica





http://www.rawstory.com/rs/2011/02/22/exclusive-militarys-persona-s...

[Trip Krant]

Who would be skewing the sentiment analysis?  Company A skewing Company B, or Company A doing it to itself?

 

What immediately comes to my mind is the potential for abuse by stock pumpers and bashers online.

[monica nixon]

Hey Trip,

Well, you have to understand how the analytics of SM monitoring systems work which I'll touch on a bit below, and that said, both scenarios you posit below are entirely possible.

So yes, the Social Media/PR Brand monitoring companies employ webcrawlers that go out and troll the web, pulling in data (comments from people, etc) across various sites/blogs , then humans "score " the gathered comments positive, negative or neutral (few of the SM companies have very effective or automated sentiment analysis abilities)-- which is critical to their being able to measure conversational tone and therefore enable their clients to effectively respond, etc. SO if some smarty decides to create false identities to skew the volume and tone broadly, uh huh how then are the SM monitoring firms going to handle this if and when it makes it's way into the private sector and say a bunch of folks who work at Bank A say Bank B's new credit card sucks, or X telecomm company is ripping off their customers, etc...you get the idea. It isnt real folks with experience with the product saying things out in the open frontier of the web, it would be folks with a vested interest, and yeah ripe for abuse, skewing of data as you wisely mention.

[Trip Krant]

Scale seems to be an issue.  You would need a lot of these persona’s (hundreds, thousands, tens of thousands?) to seriously skew the monitoring platforms, right? 

[monica nixon]

Trip,

To be able to skew the data meaningfully would depend on the volume with regard to a particular topic -so say there are 100 comments about a new cellphone that just came to market and user experience with it that the web crawlers pull in, and 50 of them are from actual purchasers of the cellphone and 50 are from the marketing dept of the manufacturer. Uh huh there you go-skew...

M

[Mark Johnson]

I'm fascinated by this news.  I wasn't aware that companies were caught using social media to do this, and it presents a great opportunity to discuss important issues around CI.  Obviously there are ethical issues raised, but I'd suspend moral outrage for a moment.

 

I'm more interested in how robust our analysis is to a single information input or a single source.  It's desirable to deliver a conclusion early, not unlike the journalist who tries to publish a "scoop" on his competition.  This must be balanced against the potential of snap judgements being proved wrong. I would beware of "Intelligence Analysis" which relies heavily on mechanized process and black box calculations.  Accuracy isn't enhanced by adding more digits to a flawed process.