>>42738990
It is more prevalent when antagonistic characters attempt to intimidate or get under my skin, but they often don't get the expected response of retaliation, fear, or anger, which tends to confuse them greatly. I usually analyze the situation for a bit before acting, unless immediately placed in harms way.
This analytical take also causes confusion when I combine it with emotional awareness. I have 'befriended' multiple changelings and antagonist throughout my chats, where I've convinced them that I was a neutral party or held some shared mutual goals.
If I do go on the offensive, I use the gathered information about the environment a present characters in creative or unintended ways; Like extinguishing all the torches in a hall with a gale of wind, knowing the unicorn guard's first instinct will be to light up their horns, which just points out where to strike to hit the horn everytime and disrupt their casting ability while leaving everyone else in the dark as I make my escape.
It just appears that the A.I. struggles with predicting my actions and is always playing catch-up. A good chunk of the chat I'll be reading about how Pony X was surprised by a new perspective, or how Pony Y was stunned by an action, while Pony Z is frustrated that they can't gaslight and often has the tables reversed. (Cozy Glow was fun for that last one, as She's likely an ENFJ)