Long-Term Memory for AI Characters: Memory, Context and Contextual Emotional Inference in AI Roleplay

An essay on long-term memory for AI characters, written by the founder of ChatBrat AI. It argues that useful AI character memory goes beyond storing facts: it combines conversational context, history and contextual emotional inference (tentative, revisable readings of conversational signals) so that remembered history changes how a character behaves. It distinguishes fact memory from experience memory, outlines what long-term memory should and should not keep, describes an observe-interpret-retrieve-compare-respond-update loop, and proposes practical tests for meaningful memory. It is an opinion and design essay, not an empirical study, and makes no product-performance claims.