The Human Controller is a scientific framework for making human behavior more intelligible and prediction-improvable. It provides a framework of the human as an energy-bounded biological controller: a body-brain system whose outputs arise from physical constraints, genetic priors, neurochemical state, prediction, attention, reinforcement history, memory, cue structure, energy, environment, and social context. Its value is practical and conceptual: it proposes a way to map behavior to the variables that generated it, rather than to a hidden chooser, fixed identity, or unexplained act of will. Thoughts, impulses, emotions, preferences, identity claims, narrative explanations, and audit responses are treated as generated outputs of the organism, not as commands from a separate internal or external controller. The framework has three proposed use layers: Human interpretation: Behavior is mapped to state, cue exposure, reinforcement history, inherited priors, energy, context, and update conditions rather than to a separate controller. This improves prediction by making the generating variables more legible, reducing blame, resentment, self-mystification, narrative defense, and inefficient error correction. Self-modeling: Indexed Identity is proposed as a structured self-representation schema for tracking goals, constraints, triggers, contradictions, state, policy, outcome, and update history. Human-AI calibration: Human feedback is treated as state-dependent controller output, not automatically stable durable preference. Requests, ratings, corrections, demonstrations, and preference reports should be interpreted through context, state, cue exposure, durability, contradiction history, reversibility, and current capacity before being stored as stable preference or used for model training. The framework does not solve AI alignment; it proposes a human-side signal model that later personal-agent, preference-modeling, RLHF, and alignment work can build on. Most component mechanisms are not claimed as new. Heritability, conditioning, predictive processing, reinforcement learning, neuromodulation, narrative identity, self-model theory, metacognition, and energy-dependent cognitive control all have existing literatures. The proposed contribution is their arrangement into one continuous controller architecture: from physical constraint and biological implementation to prediction, reinforcement, emotion, habit, identity, metacognition, and human feedback interpretation. This work is not an empirical paper, clinical protocol, self-help program, executable computational model, or completed mathematical formalization. Formal models, protocols, software implementations, and empirical tests belong to future implementation layers. Website: lawfulandpredictable.com Version 2.0 supersedes the earlier draft released as The Code v1.0.
Jan Anton Ritzl (Sun,) studied this question.