Aspirations and wants

Owns, per agent, a long-term aspiration (a goal type such as family, knowledge, or fortune), a few open want slots, sometimes paired with fear slots, and a reward stock. A generator refills empty slots by sampling wants from a pool weighted by aspiration, traits, life stage, needs, and recent events; each want is a predicate over simulation state (make a friend, gain a skill level, get promoted) with a point value. When a predicate becomes true the want pays its points and leaves its slot, and a realized fear subtracts. Points feed a meter that shifts mood or behavior and usually a currency for rewards. The Sims 2 set the form: the aspiration biases which wants and fears appear, fulfilled wants fill the aspiration meter, and an elder's remaining lifespan depends on that meter. The Sims 3 replaced it with wishes that pay Lifetime Happiness, and The Sims 4 ties a bonus trait and a goal track to the chosen aspiration. Wants do not decay like needs; unlike a career ladder they are generated from state rather than authored as a fixed sequence, and unlike villager requests they are the agent's own goals.

A stream of small, personal goals. The player reads what each character wants right now, plans the day around one or two of them, and collects a payout when they land, so an open sandbox gains a to-do list that comes from the character rather than the designer. Fears add dread: the player steers away from events the character would hate.

• number of open slots and whether fears exist alongside wants
• reroll, lock, and dismiss rules for slots
• generator weights by aspiration, traits, life stage, and recent events
• point value per want and what points buy (meter effects, reward objects, traits)
• difficulty and time horizon of individual wants relative to play pace
• whether unmet wants cost anything or simply expire

• Wants that need unavailable or off-lot actions clutter the slots and teach players to ignore the panel.
• Cheap repeatable wants let players farm points, so the reward store empties early and the meter stops mattering.
• Generators that ignore context produce incoherent wants, such as wanting a first date the day after a spouse dies, which breaks the illusion of a mind.
• Too many slots turn the panel into a checklist with no choice in it.

• Players chase wants that line up across household members, so generated goals couple separate characters' days.
• Wants seeded by recent events (a promotion, a new baby) create follow-on arcs, an informal story chain driven by state.
• Points buy rewards that make later wants easier to meet, a reinforcing loop toward a comfortable late game.

• Can a language model write wants from an agent's memories and relationships while keeping each want checkable as a predicate over game state?
• Do fears add enough drama to justify goals the player cannot fully prevent from failing?