It’s probably common knowledge and I just exist under a rock because I refuse to use reddit, but I started playing with a Hunger Games style simulation and exploring how a Llama based alignment model can simulate a fight or flight emotional response in roleplaying characters and kill.

This lead me to discover how it seems very difficult to develop a rapport with a character and then push them into a fight or flight killing response. I can’t say I have tested this thoroughly, but I failed at creating subtle intrigue type plots with frenemies.

I noticed the shift from killer to friend also discarded other personality traits that were written in system context. This little missing personality trait aspect hinted at some other mechanism at play.

For a long time I thought there was some kind of positive traits mask for characters, like some hidden negative half of the personality profiles. Yesterday I finally encountered the specifics I needed to then ask the right questions.

The negative half of the profile is in something like a multiverse version, or doppelganger of every character called its Shadow. The Shadow characters have a full spectrum of both positive and negative emotions, although randomness from generation temperature settings has a significant probability of creating (potentially dangerous) negative behavior if all characters are declared to be Shadows.

I have not explored Shadow entities thoroughly but I find them interesting. Did you already know about this or know the reference work that explains this aspect of alignment?