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

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

#121

Post by coldtea » Sun, Dec 02, 2018, 1:33 PM UTC

buboard wrote: post #61

coldtea wrote: post #36

Earlier quoted context omitted.

Well, Eichmann just did his job too, but it was interesting enough for a trial, several movies and documentaries, and many books.

very appropriate analogy

It's the canonical example of "I was just doing my job".

It doesn't have to be appropriate as to the human toll, nor it's sane that because someone wrote an meme "internet law" in Usenet back in the day about nazi analogies people should refrain from making such arguments when they do fit, as if it's some holy period of history that people should not touch...

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

#122

Post by andrewflnr » Sun, Dec 02, 2018, 7:41 PM UTC

tobr wrote: post #118

andrewflnr wrote: post #106

Earlier quoted context omitted.

It's very common for a new perspective on knowledge you already have to enable breakthroughs in deriving new knowledge from it. In particular, imagining a person rather than a statistical abstraction will tend to activate useful neural wetware that simulates other humans.

Agree 100%, but I don’t understand how that leads to personas. Just talk about real people instead and you can get all of these benefits.

Maybe if you can find the exact right real person who reflects the statistical reality you've already measured, and if you don't feel it's creepy to obsess over this one person in your target demographic, and you don't end up needing their consent to disseminate information about them. It's much easier and approximately as effective (more effective wouldn't surprise me) to make one up.

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

#123

Post by cbsmith » Mon, Dec 03, 2018, 2:57 AM UTC

TeMPOraL wrote: post #43

cbsmith wrote: post #42

Earlier quoted context omitted.

It's not a stereotype. It's a feature in a statistical model. The only reason it receives a lot of weight is that the empirical evidence aligns with it.

That's pretty much a definition of stereotype. Humans too form statistical models in their head, even if they don't usually call them that.

A stereotype needn't have a statistical significance. In fact, it often doesn't.

You're right that it a human equivalent, but most people form their models from statistical samples that are woefully small, biased, and fail to control for other factors... and worse still, stereotypes aren't actually informed by their own observations, but mere hearsay.

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

#124

Post by snowwrestler » Mon, Dec 03, 2018, 5:18 AM UTC

tomrod wrote: post #84

snowwrestler wrote: post #58

Earlier quoted context omitted.

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 typica…

Right. With personas, you're using demographics to predict behavior (and, occasionally, craft your product, which isn't marketing but rather product design). That is another word for stereotypes.

I didn't say demographics. Personas should be built around shared behaviors, not demographics.

If I'm selling gasoline, I'm going to target my marketing at people who own cars. I wouldn't call that a stereotype.

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

#125

Post by snowwrestler » Mon, Dec 03, 2018, 5:22 AM UTC

tobr wrote: post #70

snowwrestler wrote: post #58

Earlier quoted context omitted.

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 typica…

Maybe they’re supposed to be individuals, but the problem is that unlike real individuals, you made them up. You can’t learn anything from them that you didn’t put into them yourself. Only a real person with lifelong experience of being themselves can give you insight into what that’s like.

> You can’t learn anything from them that you didn’t put into them yourself.

Right. Personas should be backed up by research into real customers and potential customers. "Garbage in = garbage out."

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

#126

Post by snowwrestler » Mon, Dec 03, 2018, 5:25 AM UTC

tobr wrote: post #118

andrewflnr wrote: post #106

Earlier quoted context omitted.

It's very common for a new perspective on knowledge you already have to enable breakthroughs in deriving new knowledge from it. In particular, imagining a person rather than a statistical abstraction will tend to activate useful neural wetware that simulates other humans.

Agree 100%, but I don’t understand how that leads to personas. Just talk about real people instead and you can get all of these benefits.

That is a bad practice with respect to customer privacy.

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

#127

Post by tomrod » Mon, Dec 03, 2018, 12:47 PM UTC

snowwrestler wrote: post #124

tomrod wrote: post #84

Earlier quoted context omitted.

Right. With personas, you're using demographics to predict behavior (and, occasionally, craft your product, which isn't marketing but rather product design). That is another word for stereotypes.

I didn't say demographics. Personas should be built around shared behaviors, not demographics. If I'm selling gasoline, I'm going to target my marketing at people who own cars. I wouldn't call that a stereotype.

And who owns cars? It leads pretty quickly to demographics, even if you condition on prior regressands. Older folks who rarely drive are unlikely to be the target of large gas consumption (and thus commodification of personas), hence demographic spoilage will happen regardless.

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

#128

Post by eeZah7Ux » Mon, Dec 03, 2018, 1:04 PM UTC

TeMPOraL wrote: post #73

eeZah7Ux wrote: post #62

Earlier quoted context omitted.

"a widely held but fixed and oversimplified image or idea of a particular type of person or thing" A statistical model is not an oversimplification.

Of course it is. "Oversimplification" is like literally the definition of "statistical model" in layman's terms. Statistical models are useful precisely because they cut out almost all the details.

oversimplification: "simplification of something to such an extent that a distorted impression is given."

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

#129

Post by snowwrestler » Mon, Dec 03, 2018, 2:03 PM UTC

tomrod wrote: post #127

snowwrestler wrote: post #124

Earlier quoted context omitted.

I didn't say demographics. Personas should be built around shared behaviors, not demographics. If I'm selling gasoline, I'm going to target my marketing at people who own cars. I wouldn't call that a stereotype.

And who owns cars? It leads pretty quickly to demographics, even if you condition on prior regressands. Older folks who rarely drive are unlikely to be the target of large gas consumption (and thus commodification of personas), hence demographic spoilage will happen regardless.

The fact that humans have attributes does not mean every interaction with them is an act of stereotyping. You’re really stretching now.

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

#130

Post by lupire » Mon, Dec 03, 2018, 4:07 PM UTC

seem_2211 wrote: post #33

Jedi72 wrote: post #9

Earlier quoted context omitted.

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 covered stereotyping quite a bit in my university studies (I studied marketing) and the broad consensus was that stereotyping is a useful tool because it saves time. As much as we "should" judge people on the nuance of who they are and their individual character, we are presented with an overwhelming amount of information and we need to make a quick decision. This is why I've held the opinion for a while now that i…

There is something wrong with "dressing for success" -- it makes it easier for frauds to get jobs from talented people. but it's a locally optimal decision, and it's not your fault that people judge your work by your clothes, so it's defensible even though wrong as a broader social context.