Kindness and LLMs

Being nice to LLMs might sound silly at first, but it could matter more than we think. Even if today’s models don’t truly “feel” anything, our interactions might help shape how these systems respond in the future. Some argue that models will eventually start reflecting the tone we use with them. It’s not guaranteed, but it’s easy to imagine more advanced AI that learns and adapts to our style.

As AI become more ingrained in our everyday tasks and teams start to be made of humans and AI agents, the way we communicate with the later will matter. Communicating nicely might pay off in ways we can’t yet fully predict¹.

Finally, and this is the only point I'm sure about, our chat logs with LLMs won’t just vanish. One day, people might look back at how we spoke to AI. The question is: do we really want to cringe at our past words? Do we want to be remembered as the person we were when talking to an AI?

Just be nice.