Human Behavior Controllable Through Causal State Intervention, Paper Argues

A new paper on arXiv (2605.27580) argues that within-person variability in human outcomes is controllable through interventions targeting a dynamic latent state. The authors define state as a time-indexed weighting vector over dimensions governing how biology, physiology, and neuropsychology process events into decisions. They claim the relationship between state, decision, and outcome is causal, not correlational, and that the weighting vector is dynamic at sub-daily timescales. The research addresses a central puzzle in behavioral sciences and AI: why the same individual produces different outcomes given the same input, and why individuals differ unpredictably. The paper proposes that outcomes can be controlled in a precise, operational sense by intervening on the state and its weighting at decision moments.