Two Additional Forms of Reinforcement

This program is the most complex and least complete of any case offered here. It started as an attempt to extend the basic Matheta program to additional forms of reinforcement beyond positive reinforcement and positive punishment by including the two forms of negative reinforcement, escape and active avoidance since these two forms of behavior/learning are essential to survival. Doing so required, according to the synaptic modification protocols used here, that somehow the offset of an aversive event would cause the release of a positive fixer, thereby providing a mechanism by which escape and avoidance behavior, if successful, could increase the probability of such behavior in the future. The program below does demonstrate that capability.

What is of equal if not greater significance is the two unexpected results that “fell out” of this model as it evolved: the replication of how the opposing motives of fear and hunger powerfully influence the day to day behavior of most organisms and also how it came to be evident that Silvan Tomkins’ basic affects were implied by the program’s structure and functioning, suggesting the possibility of more comprehensive modeling of the role of affects/emotions in the behavior of organisms.

Program Description

The program is an in silico representation of a Skinner box experiment where the user presents various stimuli and rewards or punishes the behavior of the organism. In accordance with behavioral principles, this should result in predictable observable changes in the behavior of the organism over time. Again, the operation of the “nervous system” is modeled by repetitive multiplication of a “brain” matrix by a stimulus matrix which represents the continuous stream of inputs to the various neurons of the brain. Unlike other simulations, however, there are multiple interactions between neurons that must be modeled and this results in a much larger, complicated brain matrix making its functioning much harder to understand and predict. Its basic operation can be understood by referring to the figure in The Modeled Nervous System block below but some elements may need explanation.

The requirement that the offset of an aversive event would produce a positive fixer required a major re-thinking of the basic program operation. As long as the positive fixer was solely released by the presentation/consumption of food, this was obviously impossible. The solution was to have the positive fixer released not by food directly but by the reduction of hunger. If the offset of hunger released the positive fixer then it might be possible for the offset of an aversive stimulus to do likewise. The obvious solution is to postulate a neuron that does just that, represented by the C/J neuron. It is seen as a neuron that fires when inhibition is sufficiently diminished. Note that the Hunger neuron inhibits C/J and therefore when hunger diminishes, it can cause the firing of C/J. Note too that when the Fear/Terror neuron fires it also inhibits C/J, with the implication being that when a painful/threatening stimulus is removed, C/J would be released from inhibition thereby generating the positive fixer. Note also how Hunger and F/T are mutually inhibitig.

Consider the implications of the mutually inhibiting actions of fear and hunger. Both are understood to motivate behavior as indicated by the blocks below the respective neurons. But note how by postulating the hardly surprising notion that they mutually inhibit one another, a mechanism is created that replicates the get food/avoid becoming food dilemma that motivates much of the daily behavior of organisms.

It is visually obvious that the C/J neuron sits at the center of everything. C/J is the lone neuron in the model that fires as a result of disinhibition and it’s also obvious that by firing when released from inhibition, it is exactly the neuron to do what was needed to generate the positive fixer whenever food is obtained or danger subsides. However there is more significance to the meaning and function of the C/J and the F/T neuron than meets the eye but to explain it requires an understanding of Silvan Tomkins’ affect theory.

The graphic below, developed by RC, illustrates the relationship between rates and intensities of nervous system activity and a subset of the basic affects described by Tomkins. The essential point worth noting is contentment – joy is produced by the offset of the activity that produced the other affects. In the process of extending the basic Matheta model to include negative reinforcement, it became evident that the C/J neuron firing in response to release from inhibition was performing a function strikingly similar to Tomkins’ Enjoyment – Joy affect. Further, the fact that this same neuron was firing in response to a decrease in hunger and that we frequently describe the experience of satiation after a meal as being “content” led us to hypothesize that the neuron was responsible for affect Enjoyment – Joy and the release of our positive fixer. Contentment, we contend, may be fairly considered a somewhat lesser intensity version of enjoyment, and hence the “C” label for the neuron/affect.

Note that the diagram above implies the affects must also interact with each other: Interest -Excitement transitions to Fear – Terror when too intense. Distress – Anguish, a cry for help, is generated by extremes of Fear-Terror. Anger – Rage suppresses all other motives, bringing the entire capability of the organism to bear in a do-or-die fight for survival against a predator. We mention this because consideration of these issues leads the insight that neural generators of these additional affects and their associated behaviors can easily be imagined to fit into our basic framework of interacting neurons. The diagram below is a very loose, speculative schematic of such a network.

Unique Details of This Program

The most important unique element of this program concerns the firing of the Contentment neuron. For this circuit to work Contentment must fire as result of a decrease or cessation of inhibitory input. Fortunately, some neurons are known to fire upon disinhibition. Modeling that functioning within the confines of our matrix model is challenging. More creative methods may be possible than those utilized here.

The Modeled Nervous System

The Matrix Representation

The Code

Choose “AffectBrain” at link below.