A monitor-and-act unit is a physical device that has its own source of energy necessary to execute an action. It also has the property that the release of energy (i.e., the action) is being triggered by an event that occurs in the environment of the unit. A small amount of energy from the environment is sufficient to trigger a release of a larger amount of energy needed for the action. A caricature example of a monitor-and-act unit is weight suspended on a string; a small amount of energy is needed to cut the string and it triggers the “action” that releases a larger amount of energy through the free fall of the weight (see figure).
A monitor-and-act unit is the elementary component of adaptive systems. A monitor-and-act unit is characterized by the capability to interact with its environment. It obtains inputs—which is the monitor part, and it delivers outputs—which is the act part. A monitor-and-act unit may implement a negative feedback loop but can also implement other types of interactions such as a positive feedback or a one-shot action. Therefore, the concept of monitor-and-act unit is a more general one than the concept of a regulator.
Examples: A neuronal membrane at the resting potential acts as a monitor-and-act unit, energy being released when an action potential is generated. A synaptic vesicle containing neurotransmitter is another type of a monitor-and-act unit. Gene expression mechanisms act as monitor-and-act units too.
A chain of smaller monitor-and-act units can form a larger one. For example, the machinery needed to execute a reflex (e.g., patellar reflex) forms a monitor-and-act unit.
According to practopoiesis, monitor-and-act units can be organized into a poietic hierarchy.
The term practopoiesis refers to the capability of biological systems to create new monitor-and-act units. In fact, according to anapoietic (tri-traversal) theory of the mind, the process of thinking is a process of creating new monitor-and-act units.
In the terminology of reinforcement learning the entire set of monitor-and-act units possessed by an organism at one level of organization defines the policy for generating actions at that level of organization. Thus, practopoietic hierarchy is an arrangement in which, for policy A, there is a policy B whose actions change policy A, which makes it a T2-system. In T3-systems there is also policy C for changing B. TD-learning and Q-learning algorithms are then special cases of policy B.
Back to: Key concepts of practopoietic theory