Which non-neural AI model is best for simulating maximum believable human behavior (NPC)?

Question

Game development

Or what components must be present in it?
I mean AI as a separate product. The emphasis is on the player's feeling that they are dealing with something alive, so that it is difficult to calculate, but at the same time, the player must be able to use real-life skills to predict behavior, even without knowing anything about the game.
The bottleneck here is only computing resources due to the fact that there can be several AIs in one game. That is, on the one hand, the model should be very simple, like a decision tree, and on the other hand, it should be flexible enough and have signs of reasonableness.
From the conditions that a game must satisfy for AI, I decided to single out the absence of the need to recognize literary text and images. Also, you do not need to deal with animations, and in general with small details. That is, we can assume that the model of the world has already been prepared in advance in a digestible (for AI) format. For example, you do not need to recognize an apple from a picture - the data is already in the form of satiety numbers, volume, weight, expiration date, color, etc. (The question is about AI, not about the game, if that). Thus, the AI ​​deals with the macro parameters of the mechanics that are used in the game, in a convenient way, and its task is to imitate such decisions that an ordinary person would make.

Answer 1

There is a current state (what I am doing), a current goal (what I will do now after I complete the current lesson and what is required for it: conditions), energy (satiety, food, water, heat), threats (health: hunger, animals, people; energies: production in the nearest zone), monitored perimeter (visual info, acoustic info, determination of the state of the environment: threats, energies), movement speed, experience (database/decision tree based on all of the above actions), dynamic priority table (what we do now and what later, and change priorities when external factors change), a "hard tree" of possible development paths (where nothing can be changed, except for choosing a direction at a fork).
reaction speed,
periods of physiological processes (satiation time, waste disposal, moisture loss, respiratory rate, heart rate, pressure),
energy loss / recovery table for various actions (sleep, food, water),
Maybe something else ...
-----
" db/decision tree based on all listed actions" - collaborative filtering should be fine.