[__MatrixMan__]

> How far can you game a profiling algorithm?

I think pretty far. I expect the future involves nonsense layer full of AI slop being read and written by AI's. Mapping it onto the actual humans will be difficult unless you have a preexisting trust relationship with those humans such that they can decrypt the slop into your actual communications.

[alganet]

I think it's more difficult than it seems.

If I were an algorithm-profiling company I would try to anchor my profilings in the real world (what kinds of people I talk to, about what, what kinds of places I visit, what are the opinions of others on me, etc). LLM garbage would be just to draw voluntary participation in potential surveys.

It takes a particularly paranoid and stubborn individual to make the necessary efforts to consider what kinds of profiling could be done with such anchored data, and even more effort to probe it enough to acquire some knowledge on how it works.

[__MatrixMan__]

I agree that it's currently paranoid to hide your activities so that the algorithm profiling companies see you as several different people, or see the activities of millions as if they're just one... Automated misdirection on the part of the users is, so far, minimal.

But the point of such a company is to sell data on individuals, and a problematic use case for such data is to kill the ones who say things that you don't like. As that becomes cheaper and easier to do I think we'll find that it's not so paranoid to hide.

[alganet]

Maybe the endgame is honesty. Not pretending you're several people, or other convoluted ways of misdirection and disguise.

Just honestly acknowledge that profiling exists, and explicitly work against it.

That should be enough to make any algorithm company that notices something is wrong to trip on its own wires, thinking some more elaborate form of hackery/covertion is being employed.

The likelyhood of some company noticing a single user is quite low though, but if you are able to hook even a single person inside that company using nothing but honesty and no tricks, that is the best trick of all.

[__MatrixMan__]

My proposal wasn't that individuals should juggle accounts and cookies in a machine-vs-human game of cat and mouse, but rather that we should rewrite the protocols we use to play that game for us. I don't think there's anything dishonest about that--it's just that making computers lie to each other is the honest work of protecting well-meaning humans from malicious humans.

Do you have a different sort of explicitly working against profiling that you had in mind?

[alganet]

Just being honest, mostly.

I'll say things I know will get me downvoted.

I'll criticize things that could benefit me if I think they're manipulative.

I won't do alt accounts even if everyone does it.

I think most surveillance and advertisement relies on social dynamics. I attempt to play the algorithms but not the people. Sometimes it will get misunderstood, and that's fine.

[__MatrixMan__]

Ah, well I hope that sharing your honest opinion about it turns out to be an effective strategy. But I'm afraid we'll regret not interfering more directly than that.