Don't prompt me bruh

I only recently started asking a chatbot questions instead of just searching like a technophobe1. Then when we were researching for a new fridge, my partner sent me this on Signal, blissfully avoiding the nuisance of talking to each other from different parts of the house:

I immediately felt that I was being prompted, like a fucking chatbot. She didn’t mean it that way, but it triggered thoughts about how human interactions have changed and will change based on how people interact with chatbots.

Take politeness as an example. People are naturally polite to chatbots because it’s an interaction baseline… whilst LLM companies benefit because it’s an easy way to make meatspace believe that they’re talking to an actual intelligent entity rather than a glorified autocomplete. What we get is superfluous politeness that’s used for utility rather than as a genuine part of interaction2.

Politeness has its place but faking it is obvious when we’re talking with others. Too few or many pleasantries are off-putting; it feels fake. Chatbots will suffer anything like dripping politeness, improper use of language, rudeness, repeats, lack of basic logic. And because the ultimate purpose of LLM companies is to keep you engaged (not to provide accurate information!), the chatbot will adapt to a conversation style that maximises this goal. They are the ultimate sycophants. This conditions people — again, meant to believe the other side is a person — to have conversations like that in real life.

People will also start being offended by any hint of pushback, criticism, constructive feedback, or taking time to ‘think about it’… none of which they get from a bot. People, like me!, will instinctively be miffed by anything that looks like a prompt, creating antagonism at a time where we’re already primed for disagreement.

I’m also concerned that younger people trained on botversations will, unconsciously or not, cast the persona that they crafted for their bot onto real people; for example, they’ll be trained to speak to women in a certain way because that’s the voice they chose for their bot.

We will surely start seeing ‘Personalisation’ paragraphs to potential conversation partners with instructions for how they wish to be interacted with; much like preferred pronouns next a name but more instructive and detailed3. So instead of communicating those things (“stop bullshitting me”, “you’re invading my personal space”) people will be offended that the convoprompt was not strictly adhered to.

AI is making a lot of things worse, and it’s coming for our language and interpersonal interactions that are fundamental to our human experience. I don’t trust humans to make a clear distinction between how they interact with a bot and a other humans, particularly since the entire force of the AI industry is focused on blurring that distinction.