Friday, March 02, 2007

Theory of Mind

For a long time I’ve wanted to write a post explaining what consciousness is and how the brain creates it. I hope to explain in later posts the implications this theory of mind has for moral and social organization. I had hoped to limit it to 2000 words, which is about as long as a long op-ed, however I decided to include a 600 word section in which I will attempt to describe quite literally what is happening in your brain while reading this sentence.

I start with some very general definitions for concreteness which imply physicalism. From there in I will attempt to outline a theory of human conscientiousness that explains how it uses neural biology to fulfill this definition. I give special attention to the description of how emotional qualia arise through statements about physical brain processes. It could be labeled as a connectionist physicalist model (I don’t know whether this qualifies it as a reductionist or non-reductionist physicalist model however). I then briefly apply the model with the example in which I describe quite literally what is happening in your brain while reading this sentence. I conclude by outlining a few suggested further implications.

A system is function from a set of input states to a set of output states (Even more generally, I postulate a deterministic universe that can only be known with probabilistic precision in any finite time period).

A mind is a system with an ordering of inputs and that changes itself over time through a learning mechanism such that inputs relatively higher in the ordering will become more likely to occur then inputs relatively lower in the ordering.

The human mind attempts to make predictions by looking for patterns in order to better manipulate its input (senses) with its output (actions). These predictions are physically embodied by the connections between our neurons; for example the series of symbols “apple” is a set of patterns of retina input (another element of this set would be “apple”) which are closely associated with a sound pattern “aah-pull” as well as another set of patterns of retina input which is roughly circular (apple shaped really) and involves the excitation of the red spectrum sensitive neurons, and is associated with taste neurons etc… which is a tangle of neural connections. This tangle may also extend to predictions about our actions involving apples, such as if we move our hand in some way we can experience picking up the apple.

The human mind’s whole conception of objects as existing in space which they occupy in a mutually exclusive manner, is a result of the tangle of neural connections between our senses of sight, touch, and inertial balance (but mostly sight), which are connected in a pattern such that, when stimulated, eventually excites neurons connected to likely subsequent sensory input patterns (e.g. an “object” “moving towards us” is in fact a pattern of visual stimulus which will be replicated in the near future, except bigger, and which may also produce the sensation of an object colliding with us and knocking us over even further into the future). Also included in these connections are neurons associated with actions, which can be interpreted as predicting the results of actions within the context of the environment. These patterns of connections are probably formed through a combination of genetic hardwiring and learning (however, as per the above definition, the genome of a population can be understood as a mind).

I am not a neuroscientist, but my hypothesis about what is happening between a human’s ears when it is thinking is that one end of the limbic system (thalamus) sends out a signal through the cerebrum to the other end of the limbic system (hippocampus). Depending on how that signal arrives at the hippocampus end, a signal is sent back to the thalmus that either strengthens the route just taken through the cerebrum, or relatively weakens it, however, instead of traveling through the cerebrum, this return signal travels through the fornix. There are many possible paths through the cerebrum which each fork and converge thousands of times, each representing a different set of predictions associated with a certain set of inputs. This can be contrasted with the relatively simple and straight forward connections of the fornix which are largely unchanging and genetically preprogrammed. In a situation where a person can sense an apple, and is hungry, the thalamus may hypothesize wiggling some toes, by sending out a signal through the cerebrum associated with the actions that we would recognize as wiggling some toes. The pattern of connections in the cerebrum (built from past experience) will eventually tell the hippocampus that we will not experience satiation of our hunger under this course of action. The hippocampus will then send a signal back to the thalamus that either weakens the thalamus’s ability to trigger this action directly, or possibly only relatively weakens that suggestion by strengthening its alternatives.

If the thalamus then suggests picking up and eating the apple, the pattern of connections in the cerebrum should tell the hippocampus that hunger will be satiated, and the hippocampus will send signals through the fornix to the thalamus that strengthen this suggestion, possibly going through several more cycles to ensure that this course of action is correct. The repetition of the cycle may gradually excite neurons more peripherally associated with the action, bringing additional layers of nuance and complexity to bear (for example, maybe there are property law neurons which need to be gradually excited before they can complete a circuit through the cerebrum to a negative part of the hippocampus in the case our apple belongs to someone else). If we pick up the apple and successfully satiate our hunger, this sensory input also will stimulate the hippocampus in a positive, pattern-strengthening way. In this way, the thalamus, cerebrum, hippocampus loop attempts to mirror the thalamus, outside world, hippocampus loop in order to produce pattern strengthening input from the outside world.

Negative experiences (i.e. disappointment, frustration, anger) are the qualia associated with the destruction, or relative weakening of stimulatory connections in the brain. This increases the relative likelihood that alternative hypotheses or actions will be investigated or will occur relative to the hypothesis or action which stimulated the destructive chemical process. Positive feelings strengthen stimulatory connections, and extreme positive feelings are attempts to build new connections on top of existing structure (forming larger and more complex patterns) by testing new hypotheses closely related to a newly strong connection.

Because connections in the brain are strengthened by use, both strengthening and relative weakening can be accomplished through excitation. This is why I believe relative weakening may be a more appropriate description then simply weakening: a negative feeling may be a quasi-random excitation of the alternatives to a hypothesis that generated a negative connection to the hippocampus. In this way the desired result is achieved (according to the connections and stimulation of my prefrontal cortex right now): the alternatives become relatively more likely to pass into action because they are being excited and the “bad” hypothesis or action recommendation isn’t. This explains why there is a superficially similar desire to dance when elated or to throw a violent tantrum when frustrated. In the case of elation, pathways branching off the idea pattern generating the elation will be most likely to be stimulated, possibly creating new successful connections (if you get news about a pay raise you may start thinking about the way you will spend it). In the case of frustration, more basic alternatives will be explored.

A word about pain and attention: The object of one’s attention is simply additional stimulatory resources being devoted to some subset of input, whether that is external sensory input or internal reflexive “input” of a loop within the brain. A painful or annoying sensation decreases the stimulatory resources one is able to devote to the previous object of one’s attention and puts more stimulatory resources into neurons associated with the source of pain. In this way it is similar to a relative weakening resulting from a negative emotion in that it causes the stimulation of a thought pathway that is likely different from the pathways currently under attention. It also likely somehow resets the thought cycle, or disrupts it through massive spikes of stimulation (to areas not necessarily at the center of attention, though when attention and such painful stimulation coincide spectacular efficacy can result, as in the case of a fighter becoming more ferocious when injured), making large complex thought patterns unlikely to achieve proper stimulation.

A word about the prefrontal cortex, the supposed “seat of personality”: Because the hippocampus is among its most direct outputs, this area is primarily responsible for determining a series of intermediate goals to achieve some end (i.e. delayed gratification, when a system of internal rewards is necessary to organize progress towards an external confirmation, e.g. making a tool for future use). Through its excitement patterns, it can determine whether a given hypothesis or hypothetical action moves us forward, and thus should be connected to a strengthening part of the hippocampus, or whether it is a step in the wrong direction (quite literally sometimes, 1 before 2 is a good thing when counting up, but it is a bad thing when counting down) and should be connected to a weakening part of the hippocampus.

In this section I will attempt to describe quite literally what is happening in your brain while reading this sentence.

While reading the sentence, your brain analyzes the retinal input “what is happening in your brain” and connects it with another set of sensory input, images of brains and images associated with brain processes (possibly a colorful cartoonish image of a neuron with sections revealed in cut away that is from an old textbook or classroom display, or possibly energy pulses moving on some kind of circuit board grid against a gray-spongy background), hopefully this pattern associated with “brain processes” incorporates much of the preceding discussion. This is mostly done in the rear and sides of the brain where visual and auditory input are heavily processed, possibly through an abbreviated cycle that foregoes the frontal lobes initially while forming the concept “what is happening in your brain” or “brain processes” from collections of images and other sensory patterns in that area. The retinal input pattern “reading this sentence” is another set of symbols representing a collection of sensory patterns (possibly an image of a school girl reading, or a collection of language symbols on a white background with dotted lines drawn towards a cartoon eye or face,…) formed in a similar way to that of “what is happening in my brain” in that it is mainly formed in the rear and sides (and possibly top) parts of the cerebrum.

The pattern of retinal input “describe” is likely abstract enough to excite parts of the frontal lobe and possibly the pre-frontal cortex after being interpreted in the rear and sides from the senses. What is happening is that this part of the frontal area (frontal lobe + prefrontal cortex) is being wired to test for mutually reinforcing input from the two concepts “brain processes” and “reading this sentence”, i.e. a physical connection is established or enabled to the web of sensory input representing the two concepts.

The rest of this section attempts to bridge these two concepts with intermediary concepts/neurons such as “retinal input,” “connections,” the association with physical locations within the “brain” which incorporates previous knowledge, the use of the phrase “physical connection,” or “wired,” etc. which form a path leading from the concepts/neuron webs “reading this sentence” to the neuron-webs “brain processes” such that, the bridge formed in the reader’s brain by the excitation of neurons achieved through reading the rest of this section allows a stimulation of the concept/neuron-web “reading this sentence” to travel across this bridge of excited neurons to the concept “brain processes”. The connections in the frontal areas to these concepts formed by the words “describe quite literally” serves to verify this bridge formed in the rest of the section actually connects to the concepts, and not as part of this bridge between the concepts.

If you agree that I have described successfully in my attempt, your brain is making new use of this bridge, stimulating the members of the bridge’s constituent neurons to the point that neurons more peripherally attached to the bridge become excited in an effort to enrich the connections and explore implications or extension of the hypothesis (some of which may be discussed below). If you disagree, the bridge will skip steps or fall short, in which case additional attempts to create the bridge will happen as you reinterpret my words with gradually wider and more general meaning in a broadening search for a pattern. Unable to find a pattern attention may fall onto the mystery of what caused me to fail my attempt, where your prefrontal cortex will attempt to connect the information you have about how I am different from you to the information you have about the difference in our views.

The implications of this model are thus:

Identity and consciousness are the result of our memories physically embodied by our neurons and the connections between neurons. Our most essential nature is not some immaterial metaphysical “soul”, it is defined by the shape of connections between our senses and actions. Otherwise stated: physicalism.

While this model asserts that human actions are a deterministic function of environment and genetics, the concept of moral judgment and punishment are not rendered null. The potential judgment and punishment at the hands of others (as evidenced by past instances of judgment and punishment) is an important pattern of environmental input for a (properly functioning) mind in determining future actions.

This model provides a framework for extrapolating our own experience of human emotional qualia not only into other human minds but also into non-human systems that also qualify as minds. For example, an expert system which routes currency deposits to ATM machines qualifies under my definition as a mind, with its actions being its direction of currency and its senses being indications of how often ATMs must be visited to be filled, or how often they cannot fill a customer request, as well as some kind of calendar or temporal grid of reference. When a mind is happy it may still continue to change, but further changes are refinements of existing structure, filling in details or going further into the future. An ATM expert system could then be interpreted as content when there exists a large complex structure of connections within its system that remains relatively stable across time, with many weak new hypothetical structures branching off that are unstable and tend to disappear and change (and without the expansionist genetic propensity for contentment to sour into boredom that exists in humans).

The system could be interpreted as happy or elated when a new large significant relation can be added to the old stable structure, such that most of the old stable part of the structure remains an intact subset of the new structure of connections (i.e. if represented as a graph of inhibitory and stimulatory connections between a set of nodes, node A previously stably stimulatory of node B will continue that relation, and there may be no previous stable relation between node A and C, but in the new structure there is a stable relation), the bubbly effervescent nature of elation being new weak hypotheses which either are revealed or become stronger due to their reliance on the newly stable extension of the structure.

Similarly, the system could be interpreted as confused or frustrated when parts of previously stable structure become unstable due to disconfirmation. Assuming a higher currency demand in suburbs rather then urban areas on weekends, an expert system directing currency downtown on Saturday would need a major reorganization, starting at a level so basic that it is directly connected to its most primary sense: the expert system’s internal calendar. This would be an unhappy system indeed (unless it was owned by an oddly altruistic and forgiving bank). Anything that system had supposed about the connection between its calendar and other senses, such as previous ATM traffic patterns etc, would be suspect and probably invalid due to this primary error in the system’s, umm, “thinking?”.

While this model is useful for anthropomorphization of many systems, both living and non-living, it has limits. For example, the economy of a nation-state can be understood as a mind, but I am pessimistic about the utility of doing so; the individual minds of people in the economy would be just as complex and unpredictable as they are without this description as a collective mind, or assignment of anthropomorphic emotional qualia. Similarly, if in the example of a system managing a set of ATMs there was a person responsible for manually reprogramming the system rather then some artificial learning mechanism, the description as a mind would remain valid, although it would be trivially different from the human mind of the programmer.
|

0 Comments:

Post a Comment

<< Home