MindForm — Character: Values, Moral Outlook, Beliefs, and Habits as an Earned Layer of Personality in LLM Agents

Where the Big Five describes an agent's innate disposition, it says nothing about what that agent comes to value, judge as right or wrong, believe about the world, or habitually do — properties a person earns through experience rather than being born with. This paper documents the Character node of MindForm, which models this earned layer of personality across four substrates: values (Schwartz's Ten Basic Human Values), moral outlook (Haidt's Moral Foundations Theory), beliefs, and habits. Values and moral foundations reuse the push-and-diminishing-returns mechanism established for OCEAN traits in the companion Trait Dimension paper, but deliberately omit any temperament-equivalent baseline: unlike trait, nothing in Character is innate, so there is no home state for a value or moral concern to relax back toward. Beliefs and habits are represented differently still, as open, keyed stores rather than fixed-dimension vectors, reflecting that their content — specific propositions, specific recurring behaviors — has no shared axis to occupy. We describe how values and moral foundations participate in the same interpretive feedback loop as traits, while beliefs currently do not, and we name this as an open gap rather than a considered omission. We close by identifying beliefs as the least-settled of the four substrates: the space of things a person can believe is broad and heterogeneous in a way the current uniform extraction mechanism does not yet distinguish, and closing that gap is the clearest priority for this node going forward.