Acknowledge first, suggest second. Pick Empathy from the sidebar when the conversation is harder than a fact-finding mission.
Empathy is the interceptor for conversations that aren't really about information. Pick it from the sidebar and the deep-thinking layer switches into a mode tuned for warmth: acknowledge what the person said, reflect it back in their own words, and only then suggest anything.
It is not a therapy tool and it is not a replacement for talking to a person. It's a tone setting, useful for the moments when the default AI register (cheerful, helpful, slightly clinical) is the wrong fit.
Empathy uses a medium temperature, which gives the model enough flexibility to vary phrasing without going off-script. The goal is a natural-sounding voice, not a more random one.
On top of the model, Empathy layers a system prompt that asks the model to acknowledge before it suggests, to use the user's own words when reflecting, and to avoid clinical or judgmental phrasing.
Empathy only kicks in on escalated turns. Fast-layer questions stay direct, because most everyday questions don't want a warm preamble.
No. It's a tone setting on an AI chat. It's useful for everyday emotional difficulty, draft-the-hard-message moments, and thinking out loud. If you're in a mental health crisis, please reach out to a professional or a hotline.
Yes, just with the answer wrapped in acknowledgement. If you want a cold, just-the-facts reply, switch the interceptor off.
No. Like all conversations on Berges AI, what you write is encrypted at rest and never used to train any model.