Home
> Uncategorized > Cambridge Analytica
Cambridge Analytica
February 1, 2017
My newest Bloomberg post is out, in response to this article about Cambridge Analytica:
Categories: Uncategorized
mathbabe
Exploring and venting about quantitative issues
mathbabe 
Yeah, Big Data is one thing, but just the growth/force of Twitter (& some other social media) I think explains a lot. Rightwing nutjobs & Fascists have long been a component of America’s underbelly, but Twitter has promoted a previously-unavailable manipulative echo-chamber and organizing platform for them — in turn making them more politically active and bold. …And near-term, I see zero solutions to stopping them. 😦
LikeLike
I’m probably missing something obvious here, but I simply don’t understand why those that seek truth, freedom, justice, and peace for all people can’t use the same tactics and platforms to spread their messages? Surely there are much more people in the world who want those things in this life above all of this ugly talk. Maybe it is not matter of stopping those groups from using these tactics, but more a matter of the overwhelming majority of humanity in this world to overwhelm the ugliness with messages of peace. If the system can be gamed at all, then it can be gamed for good too, not just bad.
LikeLike
Can you share the link to your post?
LikeLike
I am a parent activist in the education arena and have been doing quite a bit of research into the development of adaptive online learning platforms and ties to the Department of Defense. One of the things that most concerned me about the article on Cambridge Analytica was the idea that we may be moving towards creating an online, and at times gamified, system of public education where the machines “learn” the students via facial/emotion recognition and nudge them towards desired outcomes. There is quite a bit of research being done with AR/VR simulations and tracking behavioral data within online systems. Much of this work is being done through DARPA and ADL in coordination with the Navy. Of course Betsy DeVos’ brother Erik Prince and Steve Bannon both have Navy ties. I’m wondering if you have looked into any of the ethical implications for adaptive systems that incorporate emotion sensing software? Here are a couple of links that may be of interest: http://ict.usc.edu/pubs/Intelligent%20Tutoring%20Support%20for%20Learners%20Interacting%20with%20Virtual%20Humans.pdf
http://narrative.ict.usc.edu/neurobiology-of-narrative-framing.html
https://cfwebprod.sandia.gov/cfdocs/CompResearch/templates/insert/profile.cfm?snl_id=2959
LikeLike
Shecky R, political communication is an old art. Sadly, our predictions (see the presentation at http://www.metapoll.co.uk) are leading us to believe that people decide what to vote, almost on impulse, very early on in the campaign. The campaign theatrics seem to have little, if any impact… campaigns just help to rationalise our choices based on our motives and have some impact on the turn out. It just so happens that populist campaigns e.g. Those that rely primarily on the Conflict and Joy motives, seems to prevail in (post) crisis situations similar to the ones we are now.
LikeLiked by 1 person
yer on a roll, did you see: http://www.nydailynews.com/opinion/uber-dishonest-data-dance-article-1.2961487 ?
LikeLike
WOW! Thanks for this background…
LikeLike
Indeed, same ketchup and thanks for that. But that raises question: why did Trump team do such a better job of using the ketchup the data generated? IF memory serves reported internal Clinton campaign polling was off by >5% in Wisconsin (eg). And what brilliant former google data scientist sent her to AZ and GA late in campaign but no WI (and barely MI)? Any insights? (Failure to factor in voter suppression, or what?)
LikeLike
Yes, both campaigns used big data. The difference was in how they used it. According to the NYT (Nov 12, pA20), evangelicals voted for Trump “because he stoked their fears that a Hillary Clinton administration would take away their religious liberties, use their tax-dollars to fund late-term abortions at home and abroad, and expand the rights of gay and transgender people.” The second reason is a standard right-wing lie. The other two appeal to their bigotry. White evangelicals voted 81% for Trump. Everyone else voted 60/35 for Clinton. In the wrong hands, big data is very dangerous.
LikeLike
the cambridge analytica/ kosinski stuff is like a bolt from the blue, and think its great that this stuff is seeing the light after the election, and bet there will be a lot of reverberations over it.
my current thinking on a rough model for last election that screwed up even legendary prognosticator nate silver: the campaigns have the choice of trying to mobilize existing voters, or trying to mobilize voters that havent been previously engaged as much. it looks like maybe this was an upset election due to the 2nd reason. the 2nd approach is more a long shot. clinton seemed maybe to focus on patterns of known/ established voters. it looks like trump, barring actual fraud, was able to mobilize a large demographic of voters in the 3 “upset states”. wisconsin, ohio, pennsylvania. (is that right?) ie white working class. who maybe have lower voting rates that he raised. it seems like this idea (proposed elsewhere) could explain a lot of the statistical shock that occurred.
in other words maybe the big/ classic problem with statistics showed up. “biased sampling”. the samples from previous elections were fairly well established in their samples but the sample structure/ distribution changed a lot in the last election that prior established sampling methods didnt catch/ net. the idea that most of the new voters are low tech and outside of cities could support this. its possible the prior statistical analysis is city/ metropolitan-biased.
LikeLike