I think you’re exactly correct. I worked adjacent to the advertising industry for a minute. It was wild what the algorithms could accurately predict, but it makes sense when you understand the math and a tiny bit of human neurology/psychology.
Now to be fair there are exceptions, like the TV ads that play a specific set of frequencies and an app on your phone (Facebook) that listens for that frequency to tie watching habits to your advertising id- that one is pretty crude by comparison, but is real enough. These days with most younger people consuming tv/Netflix/etc via apps on their phones it’s mostly unnecessary.
Easier just to track people by shared IPs and apps moving between networks, and similar browsing history patterns, then bucket you into hundreds of groups - geographical groups, family groups, friend groups, hobby / interest groups, professional groups, community groups, etc and watch the entire group for trends and interaction patterns with other groups, and project forward for each person using the groups they’ve been statistically bucketed into. Obviously there’s more to it, but those are the broad strokes and those techniques work shockingly well. We’re just dumb apes after all, our individual behavior patterns can be erratic and unpredictable, but groups of people, even small groups, are highly predictable.
Advertising is a crazy world. I’m not really sure how much any of us can escape it when ad networks like to jailbreak the boxes we put them in like that. Tiny flaws in your armor get exposed - and tiny flaws in each of your friends, family members, co-workers, and just people who you vibe with all compound like that.
And you have to assume nation states have very similar approaches- why monitor everything everyone does when it’s easier to categorize people into sticky groups and monitor and extrapolate behavior based on that? Hell, advertising data is cheap as hell and very for sale.
For personal protection- always on private VPN, noscript, browsing in sandboxes, and open source apps wherever possible will help manage your individual exposure greatly. But again, it’s not just your exposure that is being exploited - so unless you’re going to police the devices of literally every person you know and interact with you are going to be at least partially exposed to advertising algorithms. How much that bothers you and what you want to do about it is up to you.
Edit: a small amount of info on the statistics, if anyone is interested. My knowledge is as of about 2018, so I’m sure there’s been some developments but this is the basic principals.
As a fellow Asimov fan, I love the parallel.
But, just to play the straight-man: I don’t think we’d have that possibility with our current mathematics even if we were tracking everything that happened to everyone for hundreds of years. Too much chaos in the mathematical sense.
A good example of how this works in reality: Say a 24 year old woman is well known to the advertising agency. She recently has graduated college and so isn’t connecting to the college wifi as often. She’s gotten a professional job and is connecting to the wifi from a corporate network. Let’s say she is worried she might be pregnant, so she looks up period tracking apps on Google and reads a CNET article comparing them. She picks a very privacy focused one and downloads it and begins entering her data. Getting more worried after seeing the results, she calls her best friend. Her friend looks up her symptoms, then reads up on different kinds of pregnancy tests and recommend she pick one up on the way home. She does, then does the test, then unlocks her phone to call her mom only to see ads for baby monitors.
Was her microphone spying on her? Did the drug store taddle that she bought a pregnancy test? Did the period tracking app tip off an advertising agency and taddle? Nope, it was all in her profile. She’s at an age where pregnancies are most common. She is obviously working and out of school, increasing her chances of trying for a child intentionally. She recently read an article about period tracking with embedded ads that connected to her advertising ID. Her best friend - who her advertising ID is tightly coupled with - researched pregnancy tests.
Bam- baby ads.
But could the advertising algorithm have accurately predicted she’d have a pregnancy this week or month a year ago? No. She’s just in groups that have statistical variables attached to them that track such things and quickly adapt to changes in ‘risky’ behavior.