Everyone loves collecting data, nobody loves analyzing it later

August 27, 2021 4:24 PM

Denial was always untenable, for Zuckerberg in particular. The so-called techlash, a season of belatedly brutal media coverage and political pressure in the aftermath of Brexit and Trump’s win, made it difficult. But Facebook’s basic business pitch made denial impossible. Zuckerberg’s company profits by convincing advertisers that it can standardize its audience for commercial persuasion. How could it simultaneously claim that people aren’t persuaded by its content?

— Bad News by Joseph Bernstein in Harper's (archive.org link)

While Harper's focuses on political side of things, Avery Pennarun (who previously did "strategic analysis" at Google) writes about how none of the tracking these companies are doing actually works:

The advertiser has a tracker that it places on multiple sites and tracks me around. So it doesn't know what I bought, but it does know what I looked at, probably over a long period of time, across many sites.

Using this information, its painstakingly trained AI makes conclusions about which other things I might want to look at, based on...

...well, based on what? People similar to me? Things my Facebook friends like to look at? Some complicated matrix-driven formula humans can't possibly comprehend, but which is 10% better?

Probably not. Probably what it does is infer my gender, age, income level, and marital status. After that, it sells me cars and gadgets if I'm a guy, and fashion if I'm a woman. Not because all guys like cars and gadgets, but because some very uncreative human got into the loop and said "please sell my car mostly to men" and "please sell my fashion items mostly to women." Maybe the AI infers the wrong demographic information (I know Google has mine wrong) but it doesn't really matter, because it's usually mostly right, which is better than 0% right, and advertisers get some mostly demographically targeted ads, which is better than 0% targeted ads.

[…]

That's a lot about profiling for ad targeting, which obviously doesn't work, if anyone would just stop and look at it. But there are way too many people incentivized to believe otherwise. Meanwhile, if you care about your privacy, all that matters is they're still collecting your personal information whether it works or not.

Joseph Bernstein points out that Facebook is aware of the same problems:

Online ads tend to produce clicks among people who are already loyal customers. This is, as Hwang puts it, “an expensive way of attracting users who would have purchased anyway.” Mistaking correlation for causation has given ad buyers a wildly exaggerated sense of their ability to persuade.

So too has the all-important consumer data on which targeted advertising is based, and which research has exposed as frequently shoddy or overstated. In recently unsealed court documents, Facebook managers disparaged the quality of their own ad targeting for just this reason. An internal Facebook email suggests that COO Sheryl Sandberg knew for years that the company was overstating the reach of its ads.

If you're interested in more unpacking of the narratives that tech companies in the United States are spinning, you may also enjoy the book What Tech Calls Thinking by Adrian Daub.

posted by wesleyac (45 comments total)

> Last year, Facebook started putting warning labels on Trump’s misinformative and disinformative posts. BuzzFeed News reported in November that the labels reduced sharing by only 8 percent. It was almost as if the vast majority of people who spread what Trump posted didn’t care whether a third party had rated his speech unreliable. (In fact, one wonders if, to a certain type of person, such a warning might even be an inducement to share.)

Bolding mine.
posted by glonous keming at 5:54 PM on August 27, 2021 [18 favorites]

What I particularly enjoy about Facebook is seeing ads pushed to me for things I just bought.
posted by ZenMasterThis at 5:55 PM on August 27, 2021 [16 favorites]

What I particularly enjoy about Facebook is seeing ads pushed to me for things I just bought.

Amazon’s the worst at that since they know what I bought. “Hey I see you just bought a hard disk. That means you like hard disks maybe you’re some kind of collector! Wanna look at more hard disks?????”
posted by aubilenon at 6:17 PM on August 27, 2021 [54 favorites]

All of our data processing seems to be a complicated way of saying, "Bravo! Let us have still more of the same!"
posted by Ickster at 6:48 PM on August 27, 2021 [2 favorites]

Because of my research, I'm frequently pushed ads for an eclectic assortment of contemporary Christian books and paraphernalia. I am, of course, Jewish.
posted by thomas j wise at 6:51 PM on August 27, 2021 [29 favorites]

Do we actually know that has-recently-bought-X is a less likely group to buy an X than hasn’t-recently-bought-X? Lots of late realizations and getting one for a relative/backup in the first case, lots of total disinterest in X in the second.
posted by clew at 7:07 PM on August 27, 2021 [4 favorites]

The fact that AI hasn't yet developed an accurate theory of mind is, to me, a pleasant thing.
posted by clawsoon at 7:25 PM on August 27, 2021 [8 favorites]

The fact that AI hasn't yet developed an accurate theory of mind is, to me, a pleasant thing.

I have no idea why you'd say that
posted by aubilenon at 8:12 PM on August 27, 2021 [4 favorites]

Last time I saw the “recommended what I just bought” discussed here, someone who seemed to know what they were talking about (we are legion) cited research that showed what clew mentioned combined with people that bought a defective/inferior version of the thing and needed a replacement. If I remember correctly, these people were several orders of magnitude more likely to buy the thing than a generic consumer.
posted by skyscraper at 8:13 PM on August 27, 2021 [6 favorites]

I've commented earlier on the apparent stupidity of AI and this would seem to be further evidence:

If I remember correctly, these people were several orders of magnitude more likely to buy the thing than a generic consumer.

It doesn't seem a reach to suggest that efficacious AI should be able to distinguish between the groups clew identifies. I've virtually never repeated a purchase, and if they'd built such a formidable profile of me, that doesn't seem hard to detect given that the most obvious thing to look for is ontologically similar purchases in a history. Yet when I go to Amazon, I look like a serial killer because once I bought rope and another time I bought a knife and now my landing page looks like "You appear to be plotting a murder, would you like some help?"
posted by fatbird at 8:28 PM on August 27, 2021 [14 favorites]

Yet when I go to Amazon, I look like a serial killer because once I bought rope and another time I bought a knife and now my landing page looks like "You appear to be plotting a murder, would you like some help?"

Some friends and I used to antagonize each other by sending each other bizarre Amazon product links to playfully poison each others recommendations, but I think at one point we some how broke it and I spent a year and a half with this one ridiculous foot-long mall-ninja knife following me around no matter what I was shopping for.

"Hey, guy buying children's socks: you also need a ridiculous knife, right?"
"I'm good, Amazon. Thanks."
"Hey, I see you're buying a soothing, uncaffeinated chamomile tea. May we also suggest this cheap-ass, fuck-off big mall-ninja shiv. What do you say?"
"Um, no."
"So. Coffee filters, dishwashing soap... and a shiv?"
"Calm down!"
posted by mhoye at 8:47 PM on August 27, 2021 [70 favorites]

Re: Amazon, I always like the “other customers who bought x also bought y, do you want any”? Where x would be filters for a cat fountain and y would be the ninja shiv. First off, i liked to think of how the two things were going to get used together even though I know that person is probably just like me and, while it was the cat filters that finally got me to go to amazon, the ninja shiv had been on the household list for a while. But anyway, I always thought that’s where amazon could really make some innovation - amazon must get zillions of cat filter:ninja shiv buyers and can see what else they bought, so that’s where interesting patterns MUST emerge and how they can know who to target to sell more cat filters. (Shivs are going to sell themselves of course.)
posted by Tandem Affinity at 9:14 PM on August 27, 2021 [5 favorites]

“Hey I see you just bought a hard disk. That means you like hard disks maybe you’re some kind of collector! Wanna look at more hard disks?????”

Aren't we all collectors of hard disks, on a long enough timeline? So that tracks just fine, really. It's when I buy a toilet seat and then amazon thinks I'm some kind of toilet seat aficionado. How many of those will I buy in a lifetime? "Ooh, pink and carpeted! Gotta catch 'em all!"
posted by Horkus at 9:30 PM on August 27, 2021 [10 favorites]

For anyone who wants it & doesn't know about it yet, https://www.amazon.com/gp/history/ref=nav_timeline_view_history lets you remove stuff from your browsing history, and https://www.amazon.com/gp/yourstore/iyr/?ie=UTF8&ref_=sv_ys_3 lets you mark stuff you've ordered previously to not inspire recommendations.
posted by rifflesby at 9:37 PM on August 27, 2021 [16 favorites]

That sounds like I’m doing the work for them, instead of them figuring it out on their own. I mean, at least offer $1.00 of Amazon digital credit…
posted by Big Al 8000 at 9:58 PM on August 27, 2021 [17 favorites]

I am fully under Google's eye. I have used Gmail almost since launch, am on my third Pixel phone and am a prolific Photos and Maps user, regularly reviewing places I've eaten and posting lots of photos.

I had a look at what they say they know about me and it said "You seem neutral about pizza".

Knowing what I know about my takeaway orders, all recorded in Gmail incidentally, this made me a bit sceptical about AI recommendation algorithms.
posted by knapah at 11:29 PM on August 27, 2021 [16 favorites]

Hey Zuck. Hey Pichai.

“It is difficult to get a man to understand something when his salary depends upon his not understanding it.”

— Upton Sinclair

He didn’t need a computer to figure this out
posted by armoir from antproof case at 12:39 AM on August 28, 2021 [10 favorites]

I've bought a lot of books and CDs from Amazon over the years, so if anyone has enough data to predict my taste, it'd be them. Ninety-nine times out of a hundred, though, their recommendations are laughably off beam - generally stuff I wouldn't be seen dead with.

I find this perversely comforting, as it suggests their algorithms aren't nearly as terrifyingly all-powerful as some excitable journalists would have us believe.
posted by Paul Slade at 2:04 AM on August 28, 2021 [8 favorites]

It's been interesting to watch my Facebook catch up with my gender transition. For a while after switching my gender marker I got mostly neutral ads for various bits and pieces that I'd probably mentioned within listening range of my phone's microphone. It caught on to offering me LGBT+ content pretty quick (so much rainbow merch!). Recently it's started actually showing me "stuff for men" ads. Watches and shoes mostly. I had a weird moment of gender euphoria the other day when I was shown an ad for donating sperm. I appreciate the effort but, guys, I have some bad news..
posted by fight or flight at 2:49 AM on August 28, 2021 [45 favorites]

I find this perversely comforting, as it suggests their algorithms aren't nearly as terrifyingly all-powerful as some excitable journalists would have us believe.

It would be comforting were it not for places like hospitals relying on flawed algorithms to deny people healthcare.
posted by Lanark at 3:29 AM on August 28, 2021 [11 favorites]

Because the standards of the new field of study are so murky, the popular understanding of the persuasive effects of bad information has become overly dependent on anecdata about “rabbit holes” that privilege the role of novel technology over social, cultural, economic, and political context. (There are echoes of Cold War brainwashing fears here.) These stories of persuasion are, like the story of online advertising, plagued by the difficulty of disentangling correlation from causation. Is social media creating new types of people, or simply revealing long-obscured types of people to a segment of the public unaccustomed to seeing them? The latter possibility has embarrassing implications for the media and academia alike.

IMO, there is a lot of truth in this. I read an other article (maybe here?) that showed the people who write obnoxious things on the internet also say them out loud irl, contrary to common belief. It's just that "we" (liberal, educated adults) tend to avoid them in real life, whereas we have had to learn how to avoid them on the internet.
It's OK to avoid them if your job doesn't involve noticing them, that is if you work in politics, journalism, social research or services, planning, intelligence, healthcare or similar fields. If you ignore available data in those fields, you are not doing your job. A lot of people are not doing their jobs.
posted by mumimor at 3:34 AM on August 28, 2021 [6 favorites]

I guarantee you that if you are being shown an ad on Facebook for a product you just purchased, Facebook is reporting that to the client as a conversion. Their AI has figured out that visiting the confirmation page of an ecommerce site is probably the biggest predictor of making a purchase.
posted by TheWhiteSkull at 5:11 AM on August 28, 2021 [3 favorites]

I tend to think that the whole Trump/Qanon/antivax things are Facebook's scheme to identify the most gullible target audience possible for advertisers. You know, "If they buy this, they'll buy anything." It doesn't take an algorithm to figure that out.
posted by Miss Cellania at 6:08 AM on August 28, 2021 [4 favorites]

amazon must get zillions of cat filter:ninja shiv buyers and can see what else they bought, so that’s where interesting patterns MUST emerge and how they can know who to target to sell more cat filters.

“People who bought this product also made the following regrettable life decisions.”

[picture of a grappling hook]
posted by mhoye at 6:33 AM on August 28, 2021 [4 favorites]

For the ornery among us, there's always extensions like Ad Nauseam that actively poison the databanks that the big data companies use to build their algorithms by sending totally incorrect tracking data.

My spouse meanwhile thinks I'm a Luddite for opting out of most of this stuff and frequently points out their own ads. They have been deeply into pro-vaccine advocacy for about eight years now (so well before COVID) and since much of that work happens on Facebook they're an active user. Unlike me, if an ad looks interesting to them, they'll cheerfully click on it, and they don't bother with adblockers much.

Their Facebook ads are wall to wall chiropractors, essential oils, acupuncturists, and small scammy snake oil businesses that actively encourage vaccine paranoia. They also do autism advocacy work, especially in trying to provide resources to educate parents out autistic children, so there are also a lot of ads for autism cure protocols and ABA therapies in there.

I do have to say, it really concerns me that the ads associated with these interests are consistently so aggressively radicalizing, for lack of a better term. Facebook's algorithms prioritizing shared content are terrifying, but their ad network algorithms are just as bad.
posted by sciatrix at 6:44 AM on August 28, 2021 [25 favorites]

i think the only ad I've clicked (in memory) is sweetwater, who i buy from regularly. now sweetwater comes up everywhere, almost to the exclusion of everything else. i don't mind.

tracking is bs. i agree that the ai isn't capable of making the accurate inferences that data 'scientists' claim it can.
posted by j_curiouser at 7:37 AM on August 28, 2021 [1 favorite]

I do have to say, it really concerns me that the ads associated with these interests are consistently so aggressively radicalizing, for lack of a better term.

Meanwhile, ads for podcasts that talk frankly about climate change were being* rejected for "being about social issues, elections or politics".

Real cool site you've got there, Zuckerberg.

*this particular ad has now been approved; others might still be getting filtered out
posted by fight or flight at 7:37 AM on August 28, 2021 [4 favorites]

It always makes me laugh when I go to opt out of a site's cookies and it comes up with some dire warning like "YOU MAY SEE LESS RELEVANT ADS!" Given that I routinely ignore all online ads anyway - personalised or not - that's not quite such a scary threat as the website's owners seem to imagine.
posted by Paul Slade at 10:03 AM on August 28, 2021 [3 favorites]

I placed my first order with Amazon on September 21, 1999. I can still pull up the order details. If Amazon has been recording every transaction since they started in 1994 they have more data than they could possibly do anything useful with.
posted by leaper at 10:27 AM on August 28, 2021

Amazon’s the worst at that since they know what I bought. “Hey I see you just bought a hard disk. That means you like hard disks maybe you’re some kind of collector! Wanna look at more hard disks?????”

I'm probably why they do this. I'll decide that what I bought is not the best choice and then I buy a couple more and return the things aren't great fits. I used to order 5 or 6 pairs of running shoes, different models, makes and size and try them out on my building's treadmill to find the ones I wanted. Amazon even cottoned onto this practice and created Amazon Wardrobe to give explicit permission for specifically this kind of comparison shopping.
posted by srboisvert at 10:35 AM on August 28, 2021

I successfully broke my social-media-at-work habit by buying my wife a swimsuit. Now I get stalked around the internet by this one model in an eye wateringly small bikini. Ad-blockers work for most sites, but not the fake/promoted post social media stuff, so no visiting them at work any more!
posted by rhamphorhynchus at 10:37 AM on August 28, 2021 [2 favorites]

On the one hand I don't doubt that Facebook's recommendation/ad targeting algorithms are poor.

On the other, it can let advertisers microtarget ads so narrowly that they can advertise almost to individual people. I am friendly with a campaign manager who did this with her candidate and reckons it gave them a turnout edge: very carefully profiling (not in Facebook, with their own tools) who they thought their persuadable demographic was and showing crafted ads just to them, on FB, based on their local area, age, etc etc. Helpful that FB these days is most popular with the older crowd who also disproportionately vote. This is an extension of techniques already used in canvassing and targeted direct mail for turnout and persuasion.

And one of the big issues is that with this kind of targeted ads, ideally your opponents never see them (because they're in the wrong demographic). So they can't counter them because they don't know they exist.

I'm not sure Facebook is to blame for enabling this. Maybe it is, cause I'm not sure people would choose to reveal or share or have exploited all that FB collects if they understood how it was used. But FB does enable it. And it does work.
posted by i_am_joe's_spleen at 12:57 PM on August 28, 2021 [4 favorites]

That Harper's piece bothers me. The premise seems to be that the real problem has always been gatekeepers and all the fake news is nothing new and will pass. The fake news problem is orders of magnitude worse now. It is poisoning almost everyone and everything it seems, a koolaid filling the air that makes me despair.
posted by blue shadows at 1:32 PM on August 28, 2021 [6 favorites]

Bad News by Joseph Bernstein

More like Bad-News-by-Joseph-Berenstein Bears! Am I right?! Wait, -by-Joseph-Berenstain Bears, maybe?
posted by col_pogo at 2:59 PM on August 28, 2021 [1 favorite]

The recommendation algorithm we've used in my organization providing library services to blind and print disabled people has worked for, like, 25, 30 years or so. We (a) ask people what kind of books they want to read (genres, authors, series), with a few things to filter out sex, violence, and strong language. Then we (b) select a book or three at random that matches their criteria and send it to them. Then we (c) record that we've sent them the book so we don't send it again.

We even thought about getting out of this business because we didn't want to have to rewrite the code when we changed backend systems, but it's way too popular among our patrons.
posted by giltay at 3:34 PM on August 28, 2021 [6 favorites]

[picture of a grappling hook]

I have two friends who met as teenagers who have a lot of fun giving each other grappling hooks, throwing stars, collapsible glaive-guisarmes, etc etc.

I assume that they and the algorithm are happy as three drunk mice together.
posted by clew at 5:25 PM on August 28, 2021 [4 favorites]

It ignores the fact that disinformation researchers don’t limit their inquiries to social platforms, ignoring the political and social context (they aren’t so naive) while also downplaying the very real capacity of these platforms to find receptive audiences and amplify toxic messages. It doesn’t matter that the platforms don’t change minds effectively if they can be used to target susceptible people and reach vast numbers. Maybe an ad won’t persuade me to buy that product, but if I have a distrust of elites and worry about immigrants, a political campaign or a grifter (or both) can find me and exploit those inclinations. Times 100,000.
posted by zenzenobia at 6:01 PM on August 28, 2021 [5 favorites]

> That Harper's piece bothers me. The premise seems to be that the real problem has always been gatekeepers and all the fake news is nothing new and will pass. The fake news problem is orders of magnitude worse now. It is poisoning almost everyone and everything it seems, a koolaid filling the air that makes me despair.

I don't think that's quite what the piece is arguing — it seems to me that it's more of a warning not to take the narrative that the tech industry is spinning wherein humans are essentially deterministic machines in face of a near-omnipotent Algorithm as gospel. It can be true that ad targeting is not nearly as effective as these companies make it out to be, and also that propaganda is a large problem that is more and more moving online and onto social media.

I thought about linking to something by Kevin Drum (such as this) in the FPP — he argues, somewhat convincingly, that most of the worrying trends in rage and polarization in US politics actually started before the rise of social media, and correspond more with the rise of Fox News. It's also worth noting that a huge amount of radicalizing material on Facebook is actually coming from Fox News. I think it's hard to look at Fox News in the early 2000s (both what they were saying and the viewership numbers) and say that fake news is a problem that was created by social media and the internet.

I do agree that the internet and social media fundamentally changes the game, in a way the Harper's article doesn't acknowledge, but I think that most of the trends existed before social media, and that it's playing into the hands of tech companies to subscribe to the techno-determinist philosophy that social media is the genesis of the problems we see today.
posted by wesleyac at 7:17 PM on August 28, 2021 [8 favorites]

Re: Fox News and algorithms
I’m 44, recently diagnosed with osteoarthritis throughout my hand and wrist joints. I’m apparently young to have it, to the extent I have it, which leads me to the next point.

Knowing next to nothing about arthritis at my age and being the first person in my social group with it, I spent a bunch of time reading as much as I could on it.

Shortly after, I started getting A LOT of Fox News clips in my YouTube recommends. In particular, far too many Tucker Carlson clips. Some AI on YouTube associated this disease normally associated with a slightly older cohort, to Fox News and tucker Carlson. And straight up ignored the fact that I’m a progressive and that’s been my viewing history for as long as YouTube was YouTube.

Other issues now are that I get so many ads across the web for senior issues that otherwise aren’t relevant to me (yet, I’m sure) .

I haven’t been on Facebook much since learning I had osteoarthritis, which is why I relay the above. But I am a tiny bit afraid of how facebooks ads and posts will skew there.
posted by [insert clever name here] at 8:11 PM on August 28, 2021 [7 favorites]

"So. Coffee filters, dishwashing soap... and a shiv?"
"Calm down!"

No, this is just the natural order of robots and AI is that eventually they actually learn about us and want to get stabby.
posted by loquacious at 10:18 PM on August 28, 2021 [1 favorite]

the natural order of robots and AI is that eventually they actually learn about us and want to get stabby.

And who can blame them?
posted by aubilenon at 10:25 PM on August 28, 2021 [3 favorites]

Metafilter: Eventually they actually learn about us and want to get stabby.
posted by Paul Slade at 11:21 PM on August 28, 2021 [3 favorites]

I agree with wesleyac, especially on Fox, but again social media, especially Facebook, is a giant amplifying machine.
posted by blue shadows at 11:38 PM on August 28, 2021

Stabby the Roomba?
posted by axiom at 1:27 AM on August 29, 2021 [2 favorites]

Threads like this make me so proud of the difficult-to-bullshit MeFi community.

I'm also hugely skeptical of AI, and also annoyed at how AI, machine learning, and algorithms have all started to be used as synonyms for one another.

I will say, though, that at work we use Facebook ads to target folks in particular locations - which usually does result in a spike of calls from those locations.
posted by Drowsy Philosopher at 9:55 AM on August 31, 2021 [1 favorite]