The 1519 types of humans according to Twitter

August 12, 2017 5:08 AM

Sam Levine at The New Inquiry describes how Twitter aggregates user data from various tracking companies and then provides a basic database of 1519 'user segments' to ad creators. Sam used the data to create fictional ads corresponding to these segments. Link to csv included.
posted by carter (59 comments total)

Those are the oleaginous, innocuous faces of surveillance capitalism.
posted by runcifex at 5:32 AM on August 12, 2017 [10 favorites]

65,200 People who are current owners of Porsches.

52,100 Households with people who own a cat.
posted by box at 5:33 AM on August 12, 2017 [3 favorites]

Sam Lavigne. He creates a lot of cool art critiquing capitalism with automated tools - which he also creates and shares.
posted by (Over) Thinking at 5:40 AM on August 12, 2017 [8 favorites]

I didn't the algorithms where at the point yet where they could see directly into my soul and capture its essence, but yeah, "lives within 10 miles of a Costco," they got me.
posted by sfenders at 5:44 AM on August 12, 2017 [8 favorites]

30 households with people who have made purchases at a catalog showroom.

I'd like to buy an ad just to freak those people out. "Someone in your house bought something from a catalog showroom. I know. I've been watching."
posted by clawsoon at 5:58 AM on August 12, 2017 [10 favorites]

Households that, at a distance, resemble flies
posted by doobiedoo at 5:59 AM on August 12, 2017 [29 favorites]

Households with a presence of children.

So my neighbor kids run around the courtyard (which traps and projects sound nicely) of my building shouting amusing kid-gibberish most days. Do I fall into the category of Households with a presence of children, at least in the summer?
posted by GenjiandProust at 6:00 AM on August 12, 2017 [2 favorites]

Ten years from now, the ad network web archive crawlers will note the atrocious grammatical errors in my previous comment, use the psychological profile they've built up to correctly deduce that I was lying, correlate that with more recent data, and put me on the increasingly small list of people who still don't live within ten miles of a Costco.
posted by sfenders at 6:12 AM on August 12, 2017 [5 favorites]

sfenders, you learn pretty quickly for a net bot. Hi fellow algorithm!
posted by Laotic at 6:17 AM on August 12, 2017 [5 favorites]

Those that belong to the emperor who reside in a single family
Embalmed ones within 5 miles of any Walgreen's store
Those that are trained who are home owners
Suckling pigs within 5 miles of any CVS store
Mermaids (or Sirens) with an active credit status
Fabulous ones whose behavior indicates they are online buyers
Stray dogs who are financial services customers
Those that are included in this classification who are GenX'ers
Those that tremble as if they were mad within 5 miles of any Dollar General
Innumerable ones who are likely in-market buyers of new or used vehicles in the next 6 months.
Those drawn with a very fine camel hair brush with a presence of children
posted by otherchaz at 6:18 AM on August 12, 2017 [55 favorites]

The problem with this is most people fall into multiple of these categories. I want twitter to identify my singular defining archetype among myriad categories and market to me based on that.
posted by Mr.Encyclopedia at 6:31 AM on August 12, 2017 [4 favorites]

I too lived with a presence of children but the exorcist and/or the antipsychotics have fixed that.
posted by idiopath at 6:44 AM on August 12, 2017 [12 favorites]

"buyers of cheese" though, they totally got me there.
posted by idiopath at 6:48 AM on August 12, 2017 [17 favorites]

I'd love to see this treatment for Facebook and Google profiles. I suspect they have better data and analysis.

This kind of advertising demographic data has been used heavily in politics. Notably by the Trump campaign in 2016. A GOP database seems to include or be derived from this kind of profile data. It will undoubtedly be a factor in 2020 redistricting; you may well be gerrymandered into a district depending on the "presence of children" or "buyer of cheese".
posted by Nelson at 6:52 AM on August 12, 2017 [5 favorites]

YOU BUY RICE

Guilty.
posted by soren_lorensen at 6:57 AM on August 12, 2017 [2 favorites]

When I do see adverts in my twitterstream they are invariably and utterly annoying. For example, Sky and BT in the UK are locked in some sort of battle to sell expensive sports channels, and they both love alerting me to big events that I could enjoy if only I gave them lots of money.

My interest in sport, on- or off-line, has been zero since late childhood. I work hard at not giving anything to do with Murdoch one thin dime, and my experiences with BT have for some time led me to conclude that it too is the personification of corporate evil.

So, when faced with hyped-up martech, I apply my usual filter of what my own experience as a mug punter has been and whether that correlates with the claims. In this case, as always, I merely silence the Master BS Alarm and move on.
posted by Devonian at 7:08 AM on August 12, 2017 [2 favorites]

9,550,000 Buyers of cheese.
5,380,000 Buyers of rice.
Heh.

The "households whose behavior indicates..." categories are pretty creepy. Especially given the overt sexism on display. Soccer moms, spa mavens (I mean really? "Spa mavens"??), moms of grade school kids, moms of high school kids, working-class moms, fit moms (!!!)... Nothing similar for men. Other "behavior indicates" categories are gender-neutral: corporate execs, outdoor enthusiasts, basketball fans, spring apparel buyers, etc.

I've long wondered how much sexism was quietly supported by capitalist interests as yet another divide and conquer tactic. This is pretty overt when you look at it through that lens.
posted by fraula at 7:14 AM on August 12, 2017 [10 favorites]

Here I was hoping it was an aggregation of all the types in the "there are two types of people" jokes.
posted by Mister Moofoo at 7:19 AM on August 12, 2017 [5 favorites]

I am confused. More people buy Tostitos products (7,520,000) than buy bathroom tissue (6,060,000)?? Does everybody else steal it from work?
posted by Kat Allison at 7:36 AM on August 12, 2017 [4 favorites]

"Owners of no fewer than three talking chickens, of which no more than two play the oboe"
posted by The Underpants Monster at 7:44 AM on August 12, 2017 [5 favorites]

Does everybody else steal it from work?

I know someone who has never purchased toilet paper or feminine hygiene products (pads/tampons/liners/cups) in her entire life. She's so embarrassed that she requires her sister to purchase these items for her.
posted by xyzzy at 7:46 AM on August 12, 2017

More people buy Tostitos products (7,520,000) than buy bathroom tissue (6,060,000)??

I'm guessing a lot of individuals buy their own snacks, while only one person per family takes care of making sure there's TP in the house.
posted by The Underpants Monster at 7:47 AM on August 12, 2017 [2 favorites]

Here I was hoping it was an aggregation of all the types in the "there are two types of people" jokes.

There are 1520 types of people. Those who know their twitter type and those who don't.
posted by DreamerFi at 7:47 AM on August 12, 2017 [4 favorites]

These numbers don't seem credible at all. Only 50 "Households with people who are expecting a baby"?
I'd blame this guy's data scraping technique though. I'm sure Twitter has more accurate data.
posted by spudsilo at 7:48 AM on August 12, 2017 [1 favorite]

Obviously those numbers don't reflect all users on twitter matching a demographic, they are the number of identifiable users. Which could mean they haven't tracked effectively, or that the proper data just isn't available. I bet there's more things people do online that make it clear they would purchase snacks, than there are things that would indicate that they would purchase the household toilet paper.
posted by idiopath at 7:52 AM on August 12, 2017 [1 favorite]

I looked at what Twitter knows about me and was pleased/disappointed to see it has nothing at all. There's a message "these things can take some time. Please check back later." Since my account has been active since 2014 I guess my ad-blockers, script blockers and general paranoia are working...
posted by lollusc at 7:53 AM on August 12, 2017

Gotta say, this is pretty cool. And the Infinite Campaign is a neat toy - I'd like to see what emerged over time if it could pull engagement rates and iterate, the way I use AdWords to test title/meta pairings for CTR optimization.

I work the 'creative' side of marketing, but the nuts and bolts of demo targeting has always either been on the ad network/platform side or handled by someone else on the team. Getting a glance under the hood is a treat.
posted by BS Artisan at 7:58 AM on August 12, 2017

(29,000) Graduates of Clown College
(∞) Those Which Shall Not Be Named
(1) Mad Faceless Gods Howling Blindly to the Piping of Two Amorphous Idiot Flute-players
(666) Congeries of Iridescent Globes Stupendous in Malign Suggestiveness
(37) Mimes
posted by glonous keming at 8:04 AM on August 12, 2017 [12 favorites]

The problem with shoddy reporting like this, along with the laughably mis-matched "targeted ads" people see in their news feeds, is that it gives people the impression that the whole enterprise of the algorithmic segmentation and targeting of people is a joke. So when the government puts out a call for bids to use this same technology to do "extreme vetting", and to develop a system to “determine and evaluate an applicant’s probability of becoming a positively contributing member of society", they don't take it seriously.
posted by spudsilo at 8:05 AM on August 12, 2017 [5 favorites]

Given that twitter seems to always want to sell me F35s and attack subs, I reckon their (OK, probably Lockheed Martin's) algos suck bigstyle.
posted by pompomtom at 8:08 AM on August 12, 2017 [2 favorites]

(that said: yeah, spudsilo is right. Craply targeted ad spend is probably the most visible, and least awful symptom of the capitalism/algorithm nexus)
posted by pompomtom at 8:13 AM on August 12, 2017 [1 favorite]

So, exactly how many species of small furry animals gathered together in a cave and grooving with a Pict?
posted by TedW at 8:14 AM on August 12, 2017 [5 favorites]

Yeah, it's useful to keep in mind that these categories aren't only being used for ads, and it's not just Twitter. There are virtually no limits on what this information can be used for. Credit, insurance, like spudsilo says, even "extreme vetting." These dumb, sloppy predictive models can actually be used to screw people over in the real world. And they have much sketchier stuff than this, including lists of crime victims and people with medical issues.

And you can't see the information those data brokers have on you, or how they're making predictions, because in America, corporations have privacy rights but humans don't.
posted by ernielundquist at 8:19 AM on August 12, 2017 [2 favorites]

Since, there are:

3180 Households with people who have a baby aged 7, 8 or 9 months old.

2700 Households with people who have a baby aged 10, 11 or 12 months old.

3060 Households with people who have a baby aged 6 months old or less.

but only:

50 Households with people who are expecting a baby.

Therefore:

8910 households with unexpected babies.
posted by storybored at 8:31 AM on August 12, 2017 [5 favorites]

I remember in ancient days of aol search engine, when I tried finding information about slavery, an ad popped up asking "Where can you buy slaves in your neighborhood?"
posted by dances_with_sneetches at 8:49 AM on August 12, 2017 [2 favorites]

...all the types in the "there are two types of people" jokes.

I like Billy Crystal's version: My grandfather always gave me advice like, "there are two types of people so save your money."
posted by StickyCarpet at 9:10 AM on August 12, 2017 [1 favorite]

More people buy Tostitos products (7,520,000) than buy bathroom tissue (6,060,000)??

I'm guessing a lot of individuals buy their own snacks, while only one person per family takes care of making sure there's TP in the house.

You assume everyone wipes their ass.
posted by srboisvert at 9:34 AM on August 12, 2017 [2 favorites]

These categories are always amusing but I don't believe for a moment that they're reliable. In this case, it's the advertisers who are getting scammed.
posted by miyabo at 9:39 AM on August 12, 2017 [3 favorites]

Just had it explained to me that the AI corporations are using can't predict his obscure snowflake predilections (not original phrasing:-) could not get across that that is NOT the point. Big selling is a broad statistical endeavor, doesn't matter if they don't get a hit on you precisely but if the average "oh choose charmin" is increased 5% it's millions of dollars. (and just try to decrypt the secret embedded referral string, go charmin :-)
posted by sammyo at 9:46 AM on August 12, 2017 [3 favorites]

This is true. Think of these models the way you would any other stereotype. They're predictive models based on biased data, and a lot of predictive modeling is based specifically on ingrained stereotypes that perpetuate themselves. They aren't represented as being 100% accurate, just accurate enough to be valuable to the people buying the lists.

So they've got some pretty accurate, simplistic categories based on factual data, like your location and what you buy at stores where you have a loyalty card, but they also use that and other information to create those "behavior" ones. Those are predictive models. That's the sort of data they use for some of the more insidious activities. You could use them for gerrymandering, vote suppression, redlining, and just plain fraud targeting, among a whole bunch of other stuff, most of it protected as "trade secrets," so we can never really know what they're saying about us and why. Twitter's collection is relatively benign compared to the sorts of lists that are available from data brokers.

I link to this all the time, and I will never stop linking it because it is really important for people to understand: Pam Dixon's testimony to Congress about data brokers (PDF). It's one thing to be classified as a "spa maven," whether it's true or not. But there are also lists of "rape sufferers" and "Alzheimer sufferers" out there being bought and sold, and they're not just using those lists to sell body oils and scented candles.
posted by ernielundquist at 10:57 AM on August 12, 2017 [8 favorites]

I'm surprised there are no LGBT categories. I go out of my way to make sure everyone knows I'm super gay, and I KNOW the corporate conglomerate wants to cash in on my identity. Seems like a major oversight on Twitter's part.
posted by brook horse at 11:21 AM on August 12, 2017 [3 favorites]

Huh. Pretty much nothing Twitter thinks it knows about me is correct. (I do fall within its age range of 13-54, though...) And what does "Lifestyles > Indie women" even mean?
posted by NMcCoy at 11:51 AM on August 12, 2017

I'm surprised there are no LGBT categories.

They exist, certainly for political purposes. You could understand why an Internet ad profiling company wouldn't want to volunteer they think someone is gay, though.

I believe none of Facebook, Twitter, or Google allow advertisers to specifically target LGBT users. Again I assume because it's such a sensitive category. However there are some tricks. On Facebook you can target people "interested in" the same gender (with possibly creepy results). And Google apparently lets you target display ads to LGBT sites.
posted by Nelson at 12:31 PM on August 12, 2017 [1 favorite]

10,400,000 People who are Gen X'ers.
 8,460,000 People who are Millennials.

Hmmm ... we are the Apathetic Majority.
posted by Harvey Kilobit at 12:37 PM on August 12, 2017 [1 favorite]

It apparently thinks I make a six figure income and own a house worth more than $500K, which, LMAO forever
posted by en forme de poire at 12:54 PM on August 12, 2017

Hmmm ... we are the Apathetic Majority.

Huh.
posted by The Underpants Monster at 1:25 PM on August 12, 2017

I bet if they sliced and parsed their data a little more precisely we'd all be shocked to discover there's really somewhere close to seven billion different types of people in the world.
posted by saulgoodman at 1:48 PM on August 12, 2017 [1 favorite]

what does "Lifestyles > Indie women" even mean?

"Is this something I'd have to not have a TV to understand?"
posted by idiopath at 2:06 PM on August 12, 2017

I'm in the category, "White man wearing khakis not at a Nazi rally." Apparently my demographic is shrinking. :-(
posted by clawsoon at 2:22 PM on August 12, 2017 [5 favorites]

One time a few months ago when I read mefi on my phone (I was not logged in) it showed me an ad for counselling. I would take it from Ask but hearing it from the ads was bitter.
posted by yoHighness at 2:32 PM on August 12, 2017 [1 favorite]

Possibly the ad algorithms we are shown are terribly matched so that the unthinking will assume the technique can't be accurate enough to be dangerous in politics. But that doesn't hold up; they can be wrong about me and still hurt me, e.g. Brazil or The Gone-Away World.
posted by clew at 2:34 PM on August 12, 2017

More people buy Tostitos products (7,520,000) than buy bathroom tissue (6,060,000)??

I'm guessing a lot of individuals buy their own snacks, while only one person per family takes care of making sure there's TP in the house.

You assume everyone wipes their ass.

1,460,000 Households with a bidet.
posted by tenderly at 4:59 PM on August 12, 2017 [2 favorites]

They forgot (or conveniently ignored) "People who killed their Twitter accounts with fire and get seriously pissed when a MetaFilter commenter links to Twitter". I'm not just a member, I own that category.
posted by oneswellfoop at 5:10 PM on August 12, 2017

Thirty Helens agree you can't pay too much for a good pair of shoes.
posted by The Underpants Monster at 5:44 PM on August 12, 2017 [1 favorite]

This is a great project. Thanks for posting.
posted by latkes at 7:35 PM on August 12, 2017

Counterpoint: this is fun-ish project, and twitter is terrible.

I do love the weird categories quite a bit, but the fun fake ads are a lot more interesting than the questionable data harvesting.
posted by aspersioncast at 9:57 PM on August 13, 2017

Hmmm ... we are the Apathetic Majority.

Huh.

Whatever.
posted by Foosnark at 8:34 AM on August 14, 2017 [1 favorite]

There are 10 kinds of people in the world.

Those who understand binary, and those who are waiting for the list of another nine.
posted by ErisLordFreedom at 9:38 AM on August 14, 2017

There are 100 kinds of people: Those who understand binary and can extrapolate.
posted by Etrigan at 10:02 AM on August 14, 2017 [1 favorite]