The lack of healthy behaviors - such as physical activity and balanced diet - in

modern society is responsible for a large number of diseases and high mortality rates in

the world. Adaptive behavioral interventions have been suggested as a way to promote

sustained behavioral changes to address these issues. These adaptive interventions

can be modeled as closed-loop control systems, and thus applying control systems

engineering and system identification principles to behavioral settings might provide

a novel way of improving the quality of such interventions.

Good understanding of the dynamic processes involved in behavioral experiments

is a fundamental step in order to design such interventions with control systems ideas.

In the present work, two different behavioral experiments were analyzed under the

light of system identification principles and modelled as dynamic systems.

In the first study, data gathered over the course of four days served as the basis for

ARX modeling of the relationship between psychological constructs (negative affect

and self-efficacy) and the intensity of physical activity. The identified models suggest

that this behavioral process happens with self-regulation, and that the relationship

between negative affect and self-efficacy is represented by a second order underdamped

system with negative gain, while the relationship between self-efficacy and physical

activity level is an overdamped second order system with positive gain.

In the second study, which consisted of single-bouts of intense physical activity,

the relation between a more complex set of behavioral variables was identified as a

semi-physical model, with a theoretical set of system equations derived from behavioral

theory. With a prescribed set of physical activity intensities, it was found that less fit

participants were able to get higher increases in affective state, and that self-regulation

processes are also involved in the system.