Optimal control of psychological processes: a new computational paradigm

In this paper, we use general mathematical-statistical theorems to prove that developmental processes must be studied at the intra-individual level. We demonstrate how to model intra-individual variation using single-participant time series analysis with time-varying parameters. We use advanced signal analysis techniques based on nonlinear state-space modeling to present simulation results obtained with a new Maximum Likelihood technique based on Extended Kalman Filtering with Iteration and Smoothing (EKFIS) embedded in an Expectation Maximization (EM) loop. After showing how EKFIS results yield state-space models with time-varying parameters, we then couple EKFIS to recursive optimal control techniques to produce a receding horizon feedback-feedforward controller. In this way, we obtain a flexible on-line computational paradigm with which we can optimally control observed behavioral processes for an individual person in real time. We will present optimal control techniques using simulated data and outline preliminary applications to real time patient-specific treatment of type I diabetic patients and asthma patients.