Cambridge Analytica Used Fashion Tastes to Identify Right-Wing Voters • Hacker News Board

Re: Cambridge Analytica Used Fashion Tastes to Identify Right-Wing Voters

#51

Post by drdeadringer » Sat, Dec 01, 2018, 7:51 PM UTC

I remember a scene from "NYPD Blue" between Sipowicz and his adult son [who wanted to become a cop like his father]. The advice was that the "most basic" thing to remember was "People, Places: The things they do, the times they do them".

If you turn a hard Left at Crazy, that's profiling ... but it's not wrong. Going with the example provided in the scene, it's good to figure out that the night-shift worker at the toy factory is bringing home a toy gun for his kid at 3am Christmas Morning instead of a high-profile over-armed robbery in an otherwise deserted industrial district in the middle of the night during a holiday.

It seems like the "grey line" shrinks and expands at will depending on the context one desires.

Re: Cambridge Analytica Used Fashion Tastes to Identify Right-Wing Voters

#52

Post by whoopdedo » Sat, Dec 01, 2018, 7:53 PM UTC

Once again demonstrating that "deanonymized" data is anything but.

Re: Cambridge Analytica Used Fashion Tastes to Identify Right-Wing Voters

#53

Post by threatofrain » Sat, Dec 01, 2018, 7:55 PM UTC

carbocation wrote: post #8

everdev wrote: post #4

Unfortunately, stereotypes have utility. No one likes being caught up as a false positive in a stereotype, but when you look at mass market politics / advertising / marketing they cater to charactures of their intended audience because for the most part it seems to work. Every once in a while someone takes it too far and a group gets offended, but let's not pretend that Cambridge is the only one doing this.

Stereotype is probably the correct word, but it's overloaded here. Stereotyping someone by something that they did not choose is a gateway to prejudice: skin color, eye color, place of birth. Stereotyping someone by something that they did choose may be very useful for making inference about them, because it is a revealed preference: clothing tastes, food preferences, etc.

Why is one prejudice and the other inference?

Re: Cambridge Analytica Used Fashion Tastes to Identify Right-Wing Voters

#54

Post by hellbanner » Sat, Dec 01, 2018, 7:57 PM UTC

autokad wrote: post #11

everdev wrote: post #4

Unfortunately, stereotypes have utility. No one likes being caught up as a false positive in a stereotype, but when you look at mass market politics / advertising / marketing they cater to charactures of their intended audience because for the most part it seems to work. Every once in a while someone takes it too far and a group gets offended, but let's not pretend that Cambridge is the only one doing this.

one interesting one was that I took 5000 twitter profile photos and averaged the pixel values by male / female. unsurprisingly, the male average had short hair while the woman long. but what did surprise me was that the shirt area of the males had a blue hue and the women had a red one. its not like anyone in particular was wearing red/blue by sex, but on average they were.

Source?

Re: Cambridge Analytica Used Fashion Tastes to Identify Right-Wing Voters

#55

Post by dvt » Sat, Dec 01, 2018, 8:03 PM UTC

As much as I think Trump's idiotic anti-media rhetoric erodes our democracy, when I see the NY Times publishing articles like this, it makes me cringe. First of all, it's a complete non-article -- marketing firms used correlative data to make inferences about political affiliation? No way!

Second of all, instead of being a smart and pointed analysis as to why this might be bad, it somehow draws political conclusions from a fashion conference and displays a large header featuring the green-haired Christopher Wylie making fantastical claims about "Bannon's insurgency" -- even an iota of research would clearly show that today's "alt-right" movement was a decades long project started by Newt Gingrich (among others in Republican leadership) in Georgia, escalated during Clinton's impeachment proceedings, and fully developed during the Tea Party movements that preceded Trump.

But of course, these are complicated political machinations that have many moving parts. Instead, the NYT insults its readers by pandering technically- and politically- incompetent drivel like “Fashion data was used to build AI models to help Steve Bannon build his insurgency and build the alt-right” -- give me a break.

Re: Cambridge Analytica Used Fashion Tastes to Identify Right-Wing Voters

#56

Post by chrischen » Sat, Dec 01, 2018, 8:03 PM UTC

everdev wrote: post #4

Unfortunately, stereotypes have utility. No one likes being caught up as a false positive in a stereotype, but when you look at mass market politics / advertising / marketing they cater to charactures of their intended audience because for the most part it seems to work. Every once in a while someone takes it too far and a group gets offended, but let's not pretend that Cambridge is the only one doing this.

it is useful for the self interested.

The game theory but is that if some stereotypr, they will come out ahead in the short term.

If all stereotype then everyone loses due to the stereotype killing off any that buck the stereotype. That is, the stereotype forces its own truth, which may be bad by killing diversity and adaptation.

Obviously allowing some to do it to get ahead in the short term will inevitably lead to everyone doing it, that’s why we should discourage all from doing it.

Re: Cambridge Analytica Used Fashion Tastes to Identify Right-Wing Voters

#57

Post by iscrewyou » Sat, Dec 01, 2018, 8:12 PM UTC

dvt wrote: post #55

As much as I think Trump's idiotic anti-media rhetoric erodes our democracy, when I see the NY Times publishing articles like this, it makes me cringe. First of all, it's a complete non-article -- marketing firms used correlative data to make inferences about political affiliation? No way! Second of all, instead of being a smart and pointed analysis as to why this might be bad, it somehow draws political conclusions…

These news don’t make a big splash when other companies use profiling to identify people. This one did because those profiles were used to incite violence. It’s news, just not to your liking.

Re: Cambridge Analytica Used Fashion Tastes to Identify Right-Wing Voters

#58

Post by snowwrestler » Sat, Dec 01, 2018, 8:14 PM UTC

Jedi72 wrote: post #9

everdev wrote: post #4

Unfortunately, stereotypes have utility. No one likes being caught up as a false positive in a stereotype, but when you look at mass market politics / advertising / marketing they cater to charactures of their intended audience because for the most part it seems to work. Every once in a while someone takes it too far and a group gets offended, but let's not pretend that Cambridge is the only one doing this.

It was hilarious as an engineer who needed to learn marketing 101 for a startup, its pretty much just stereotyping your customers. "Wendy is a 24-30 yr old professional, slightly insecure, straight, likes wine and dogs."

I think you're talking about using personas. I'm not surprised that was your reaction, as personas are often misunderstood, even by marketing folks.

The point of using personas is not to stereotype your customers but to establish focus sufficient to permit clear product decisions. In other words the idea is not "all our customers are like this" but "this is the type of customer we care about the most." Or more typically there are multiple personas, which allows you to create a matrix of product features with personas and again, establish priority for which feature to develop when, and how much to invest.

I'm using "product" and "feature" here broadly, so for example even in an ad campaign, each ad can be considered a "product" with "features" like where it appears, what's in the imagery, what the text says, etc.

You can run into trouble with personas if you start believing that they cover the entire range of potential customers. They don't, and again, that's the point: focus. That's why personas are sometimes given names and photos and detailed descriptions, to make it clear they're supposed to be individuals not stereotypes.

Re: Cambridge Analytica Used Fashion Tastes to Identify Right-Wing Voters

#59

Post by inetknght » Sat, Dec 01, 2018, 8:25 PM UTC

WillPostForFood wrote: post #47

inetknght wrote: post #16

Earlier quoted context omitted.

Stereotyping someone by something that you assume they chose is also a gateway to prejudice. @NeedMoreTea said it right by mentioning cases where someone might not have chosen their particular attire.

It also begs the question of whether we have any more free will in fashion choices than we do skin color.

I can see an argument for that but I don't think it's a valid assertion. People can change opinions. They can't so easily change their skin color.

Re: Cambridge Analytica Used Fashion Tastes to Identify Right-Wing Voters

#60

Post by extra88 » Sat, Dec 01, 2018, 8:29 PM UTC

hellbanner wrote: post #54

autokad wrote: post #11

Earlier quoted context omitted.

one interesting one was that I took 5000 twitter profile photos and averaged the pixel values by male / female. unsurprisingly, the male average had short hair while the woman long. but what did surprise me was that the shirt area of the males had a blue hue and the women had a red one. its not like anyone in particular was wearing red/blue by sex, but on average they were.

Source?

They are their source:

> I took 5000 twitter profile photos and averaged the pixel values by male / female.