This is just a shower-thought/thought experiment.
It’s obvious now that bot farms are capable of pushing a narrative through brute-force repetition. But even knowing this, it’s nearly impossible to avoid the “propaganda engine” or “the algorithm”, or whatever people want to call it. It seems even casual use of social media is enough to drastically change how you view the world. We’re all falling prey to it (sadly), and it’s almost always for the worse… seeding negativity or distrust, being needlessly argumentative or unreasonable, pushing an agenda. Without getting too philosophical about the concept of free will, this seems undesirable even just on a pragmatic level.
Recently while doing some LLM-generated Anki flash cards for learning French, I had a card that I had to pause and think not only about the language, but whether it was true: I think Claude gave me propaganda! 😂 (To be clear, I believe this was inadvertant! Not implying any malice.) I’ll skip the specific example to avoid needless politics here, but it was equivalent to a card reading “The sun rises from the west every morning”. You repeat it in French and pat yourself on the back for the translation and then pause and go, wait, that isn’t correct!
So I thought, if I’m generating language cards that I’m going to repeat thousands of times in aggregate, should I use this opportunity to self-propaganda my brain? Train myself to psychologically lean in a certain direction? For example, instead of just “generate examples of the following verbs in various conjugations”, add “with a subtle slant towards the beauty of nature and positivity in the natural world”.
Probably the same way that reading repeated propaganda in social media over and over is bad for our psyche, it would be better if instead of translating “she was surprised at how expensive her rent was” or “she was surprised by the violence in the street”, we instead learned the word “surprised” by repeating “she was surprised by the beauty of the trees as the sun rose”…