3 ms·

The whole article reads, to me, as "We can sell you ads on the most tenuous associations" as we're too stuck in our own marketing bubble. :) Stereotypical diff

The whole article reads, to me, as "We can sell you ads on the most tenuous associations" as we're too stuck in our own marketing bubble. :) Stereotypical difference between a peace activist and Wall St banker, sure. No doubt some more subtle variations too. Comparing mainstream brands? That seems a stretch too far to say the least. Preferences in clothing and music are the leading indicators of political leaning Or might indicate nationality and age group far more than politics? Music has long heavily influenced clothing. opting for Levi’s over Rag & Bone makes a statement about associations and history and opens one up to fashion profiling Or you were born in the 50s and haven't, yet, heard of Rag & Bone, or you prefer Levi's fit, or they're heavier denim. Or you're British and prefer Rag & Bone, as they started here. Or one was on offer at Amazon. Or your partner chose whichever they prefer on you.

inetknght 8y ago
I don't think it was being used as the sole metric. Rather, I think it'd be used as a weight. X percent of the population wears A brand, while Y population wears B brand. What correlations can we find in some Z count of reference population? Use that correlation value to infer data where none existed. Yes, some of that inferred data will be wrong. But a lot won't be. And when you're talking a massive numbers game like millions of voters then you're also talking about accepting that trade-off by hedging against other correlations that can help solidify a potential inference.
- brazzledazzle 8y ago
  I think you’re spot on. I have friend that probably defies nearly every major category he’s part of: age, race, population density, location, home and perhaps even vehicle. But after reading this I realized his clothes do stand out and might be one of the few outward things that don’t suggest “middle aged midwest conservative”. If you need to figure out where to focus canvassing and marketing in a political campaign it’s probably not bad to have a range of variables and metrics to weigh, especially when things are otherwise mostly homogeneous. My friend isn’t going to tell his neighbors he voted for Clinton but you better believe he did.
- NeedMoreTea 8y ago
  In those terms it sounds perfectly reasonable. Just as "Preferences in clothing and music are the leading indicators of political leaning" is rather different to "Preferences in clothing and music are two strong indicators, of the dozens of factors we used to infer political leaning". Just like Google use hundreds of factors, with probably a few more significant ones, to decide search placement. I'm sure inferences can be made in aggregate, but the precision that is implied in the article, from just two indicators is a leap. Shoddy reporting, or a little too much hyperbole from someone giving a conference speech I couldn't say. It's the sort of thing that gives rise to 30 minutes of comedy in the office when there's been some reason, like a news piece, to view FB, google or other profiles. We inevitably find that half the office is mis-gendered, aged, those with children don't and vice versa, and on. These profiles seem to get most people wrong much of the time, but be worth something in aggregate. :)
lsc 8y ago
Pretty much. I think this is also why self-driving cars are still not mainstream, while we all carry little cameras in our pocket that can literally translate text in real time. Like... the things we have that work? they work 95% of the time, and for translating menus? that's amazing! and really useful. Advertising is something where getting it right 95% of the time is way overkill, too. so you can advertise on super tenuous links, and it's fine! totally works. working 95% of the time just isn't okay for self driving cars, and that is what makes the problem so hard.
coldtea 8y ago
>opting for Levi’s over Rag & Bone makes a statement about associations and history and opens one up to fashion profiling Or you were born in the 50s and haven't, yet, heard of Rag & Bone, or you prefer Levi's fit, or they're heavier denim. Or you're British and prefer Rag & Bone, as they started here. Or one was on offer at Amazon. Or all these objections just apply to this or that individual case, but in aggregate the observation is correct.

inetknght 8y ago

I don't think it was being used as the sole metric. Rather, I think it'd be used as a weight. X percent of the population wears A brand, while Y population wears B brand. What correlations can we find in some Z count of reference population? Use that correlation value to infer data where none existed. Yes, some of that inferred data will be wrong. But a lot won't be. And when you're talking a massive numbers game like millions of voters then you're also talking about accepting that trade-off by hedging against other correlations that can help solidify a potential inference.

brazzledazzle 8y ago
I think you’re spot on. I have friend that probably defies nearly every major category he’s part of: age, race, population density, location, home and perhaps even vehicle. But after reading this I realized his clothes do stand out and might be one of the few outward things that don’t suggest “middle aged midwest conservative”. If you need to figure out where to focus canvassing and marketing in a political campaign it’s probably not bad to have a range of variables and metrics to weigh, especially when things are otherwise mostly homogeneous. My friend isn’t going to tell his neighbors he voted for Clinton but you better believe he did.
NeedMoreTea 8y ago
In those terms it sounds perfectly reasonable. Just as "Preferences in clothing and music are the leading indicators of political leaning" is rather different to "Preferences in clothing and music are two strong indicators, of the dozens of factors we used to infer political leaning". Just like Google use hundreds of factors, with probably a few more significant ones, to decide search placement. I'm sure inferences can be made in aggregate, but the precision that is implied in the article, from just two indicators is a leap. Shoddy reporting, or a little too much hyperbole from someone giving a conference speech I couldn't say. It's the sort of thing that gives rise to 30 minutes of comedy in the office when there's been some reason, like a news piece, to view FB, google or other profiles. We inevitably find that half the office is mis-gendered, aged, those with children don't and vice versa, and on. These profiles seem to get most people wrong much of the time, but be worth something in aggregate. :)

lsc 8y ago

Pretty much. I think this is also why self-driving cars are still not mainstream, while we all carry little cameras in our pocket that can literally translate text in real time. Like... the things we have that work? they work 95% of the time, and for translating menus? that's amazing! and really useful. Advertising is something where getting it right 95% of the time is way overkill, too. so you can advertise on super tenuous links, and it's fine! totally works. working 95% of the time just isn't okay for self driving cars, and that is what makes the problem so hard.

coldtea 8y ago

>opting for Levi’s over Rag & Bone makes a statement about associations and history and opens one up to fashion profiling Or you were born in the 50s and haven't, yet, heard of Rag & Bone, or you prefer Levi's fit, or they're heavier denim. Or you're British and prefer Rag & Bone, as they started here. Or one was on offer at Amazon. Or all these objections just apply to this or that individual case, but in aggregate the observation is correct.