Artificial Consciousness/Neural Correlates/Functional Models/Self

From Wikiversity
Jump to navigation Jump to search


If you thought I was doing violence to the ideas of Executive Function, Will and Volition, I imagine you are going to think I have really gone off the rails with this one. Self-hood is merely a programming trick in a self-programming language.

Essentially what self-hood does is create an environment where an automatic programming language is less likely to stress the constraints of the organism it is based on. The strategy is simple, it is called Model then Act. the idea is that you want to predict how a program will operate before you execute it. So you block execution at a low level and run the model, predicting the operation of the program, and testing to make sure it doesn't violate any of the constraints of the organism. If the model works, then you retrigger it without the constraints, and monitor the real world to see how close the model came to the real thing. To help with this, you need a logging mechanism so that you can rewind any sections of the process that get out of control.

In the bicameral mind, Model then act, works really well because the Right side of the brain tends to allow more accurate models, while the Left side of the brain is more selective. It is partly through modeling that the Dorso-Lateral PFC is given enough information to select from among strategies. However this information is not generated in the DorsoLateral PFC, because it has no connection to the limbic system. Instead, the model then act mechanism is probably controlled by the OrbitoFrontal PFC area. One of the model then act models, is something called the Self-image. What it seems to be is the set of constraints that model what you are likely to do, and the source of the Self Feeling.

However this mechanism is not complete unless it also has a feedback loop from the real world, that feeds it the constraints of the body. Some of these constraints become presets, that are taken from the environment or state of the organism at certain set "sweet spots" during development. An example that is really good, is the "Attachment" stage that sets the social expectations presets for the self image. The self image acts as a governor limiting the extremes of behavior and setting them to fall within limits that are socially acceptable by using the mothers impact on the babies emotions to set the parameters. This is what attachment does. It teaches the self image how to control the emotions and other social expectations. However childhood disease at about age 1.5 to 2 years can unbalance the attachment period, and retrigger growth of the opposite lobe in an attempt to compensate, but there is no guarantee that the second opportunity for attachment will take. Experiments with monkeys that had their attachment periods blocked by lesions of the OrbitoFrontal Lobe, showed that in a great number of the cases, the secondary attachment periods were missed due to development in other brain areas getting out of sync with the re-attachment of the baby. For example, If the baby was going through a particularly active period the mother couldn't re-attach.

If this is as important to the operation of the brain as it seems to be, then the blocking signal that is removed during volitional acts, is there to allow the model then act process to evaluate the act before it executes. The extended time differential that Libet claims he has detected, might be the result of the model running first before the act. The primary difference between intention and volition, might only be that volition was programmed using an automatic programming language, and the result hadn't been tested yet to see how it fit with the constraints of the organism. While intentions are well automated, and thus can be trusted to have their parameters published and known.

A Compilation step that involves a number of Model then Act implementations over time, might be necessary to explore the constraints of the volition well enough that model and then act can be dispensed with and it can be automated into an intention. The fact that the OrbitoFrontal PFC is already involved with this type of activity makes it the candidate of choice for the controller that monitors the more general case of Model then Act, in volition.