rashbre central: anything hiding in the mechanical turk?

Monday 19 March 2018

anything hiding in the mechanical turk?


I see facebook may be about to receive a comeuppance as a result of the way its users had data processed by a Cambridge Analytica subcontractor.

My take is that a pivotal point seems to be the difference between academic and commercial research. If GSR/Global Science Research asked for data to further 'academic research' it appears unconstrained in ways that no commercial entity would be allowed.

The separate compartmentalising of the main players is another interesting facet. If everyone becomes firewalled subcontractors it is far easier to issue denials, "Not us, guv."

Although, come to think of it, the immediate issue of denials seems to be a first response to everything in these spin managed times.

I've never liked facebook. Zuckerberg's first version of face mash was a sophomore project to visually compare on-campus women. He was almost expelled for that, with privacy being one of the citations. Maybe that early disregard for privacy has persisted? You can tell I'm one of the people that think of Facebook as abusive by design right from its throw sheep and poke days.

One of facebook's recent patents is about 'socioeconomic group classification based on user features' - fundamentally another way to profile users so that they receive the right advertising. The basics are not new but there's a scrabble now to be one of the main providers.

Throw Breitbart, Steve Bannon, Robert Mercer/Renaissance Technologies hedge fund/Cambridge Analytica into the mix and the ability to run informational dominant psychological operations (psyops) starts to magnify.

That's the premise of at least part of the electoral influence. Psychographic messaging, derived from an analysis millions of facebook users. Plonking messages most likely to influence in front of voters. Politics downstream from culture.

Of course the technique is not legal, but everyone in the chain has for many months denied any part in it.

The Global Science Research data harvesting via a paid user personality test was cunning. Ask each user if it will allow the user's friends to be used to improve the data collection quality.

Thus grab data from an average of 160 friends per paid respondent capturing 50 million profiles. And maybe use some inexpensive manual help to massage the data?

But of course everyone is denying any part in it.

There's probably insufficient data.

No comments: