Monday, 15 May 2017

you are a micro target


Tammany Hall vote-stuffing exemplifies dubious voting practices. However, instead of visiting multiple polling stations or being offered booze and a shave before revisiting the same location, the fraudulent systems have become more sophisticated.

Although even as recently as last September, it was blatant vote-stuffing, such as in Putin’s Russia. Turn up and empty a handbag of extra votes into the box. Or maybe bring a line of soldiers to vote out of area.

Information Technology and Big Data have made it all more clandestine. Hack the voting machines? Use Big Data to micro-target electorate with appropriate messaging?

Recent stories feature the Brexit vote influence from Cambridge Analytica (via its –ahem- non-subsidiary AggregateIQ).

Behind this chain of influence appears to be Robert Mercer and Steve Bannon, both showing form with Trump with organisations potentially involved in the tweaking of the US electorate vote through micro-messaging. There's other interesting names to link into this, but time will out them.

UK's Vote Leave spent almost £4m with Aggregate IQ, and its spokesperson Dominic Cummings extolls their virtues on their web-site’s home page. There's also curious funding associated with all of this - I wonder how much it costs to tip a ballot nowadays? It wouldn't be a huge sum to a billionaire or a state with a particular agenda. Wanna customise America? Or maybe break up Europe? Align your favourite deck of political leaders? Perhaps there's an evolving menu of prices?

To get the information to be able to work out the profiles of the electorate requires a few Groupby-Before-Join queries, to blend various data together. The smart propeller heads can make it all seem more complicated and talk about grid-based n-dimensional parallel sparse matrix-vectors and so on, but it's really like creating a whacking great mailing list.

Cambridge Analytica has access to Facebook behavioural data, so a blend of that with, say, mailing list magazine subscriptions and perhaps the electoral register would provide a rather interesting way to classify people.

I don't know how much of what I speculate is strictly legal, mind you, but I doubt that the Data Protection Act will get in the way.

So think carefully before completing the next Facebook meme to see whether you are more like a unicorn or a rainbow.

Of course, I’m making up the exact approach used, but anyone that has used mailmerge with mailchimp filters can see how this could work.

Admittedly the number of rows of data become quite large, but nothing that can’t be handled with some divide and conquer techniques.

Although, from time to time, you can see how it goes awry if you get one of those apologetic emails from a mailshot that shouldn’t have been sent.

So, where does this leave us all?

Disenfranchised and manipulated, maybe?

No comments: