Neural Networks Explaining Econ & Human Emotion?

In the process of taking my mind off of too many things to count and needing a few minutes break from work, a student post about neural networks and artificial intelligence got me thinking about its application to today's marketplace. Everyone seems to be trying to come up with a theory to explain what is happening today; I am convinced none exists. It is a series of mistakes, blunders, oversimplifications, overcomplications and greed that got us where we are today, but perhaps it is the use of information technology - and not in the systems sense - that could get us out of it.

Technologists have long studied artificial neural networks, or ANNs.. these are often referred to just as neural nets. This is one of those terms that students are often baffled by; but its really simple when you break it down. We can use many analogies; the brain, IT, math.. If we look at the brain, we have a series of neurons.. these aren't the artificial type however. They connect information through synapse connection; in IT this is referred to as a connectionist approach. But the key about this system is that it is highly adaptive; information is constantly exchanged and new networks are created.

Neural nets are used in statistical data tools too; particular those creating models or trying to explain what isn't explained by a regression equation. We can use them to find patterns, to find relationships - perhaps even from an emotional perspective, to explain relationships. Some nodes in this network are those which are inputs... from a human perspective we have all needs on the hierarchy.. from a neurology perspective we have the senses, and so on. They also consist of those nodes which are hidden; in emotions perhaps intentionally; in neuroscience perhaps because they are nothing more than transmitters to create a synapse connection. Then we have outputs; the results of the input nodes and the hidden nodes; this works much the way a network does. In neural nets though, interestingly, only those hidden nodes produce output - one could say that only our hidden motives or internal needs are those that produce change. In a capital market, one could say that this is yet another proof of Smith's Invisible Hand; with human motivation moving markets.

One thing is certain and that is every discipline uses neural nets in their own way but I don't see it being used in economics much; or in emotions and emotional intelligence (EI) analysis much either. We do generally agree that it is taking the simple to create a complex pattern of behavior; the simple human need to be loved to create a complex set of outcomes that is often unexplainable; the brains need for food to take drastic or creative measures to get it; the markets need for profit to make uninformed and albeit seemingly irrational decisions to achieve it.

I was telling my students earlier that one goal of scientists in neural network study is to not necessarily make the network adaptive.. it does a pretty good job of that on its own. But, to be able to measure the weight of each connection to either modify it to produce an intended effect, or weigh its value. This can be used for a variety of purposes; but perhaps explaining emotion and economics (and the connection of the two even) is of most interest to me. If we can modify the inputs or if we can identify the hidden nodes, then perhaps we can explain and even give value to the outputs; this in turn could potentially predict when markets are overreacting, overrationalizing, or even overcorrecting. We can then tweak the value of the output, which would in turn create a new method of explaining what economists try to dumb down to supply and demand.

Another fascinating element is bio neural nets... with biological NNs, each of the nodes functions in units but in parallel; each node appears to understand the motivation and the hidden nodes and then react accordingly. Tasks are clearly distinguished; each node knows what it needs to do and its relationship to other nodes. Perhaps just as in the human brain, this is missing in our marketplace. Perhaps the goal to keep the market competitive has removed the ability for nodes to work in parallel, thus preventing an intended purposeful and intelligent outcome.

Scientists like systems that adapt; after all if we can tweak the input we can tweak the output. But what seems largely missing in most disciplines is the hidden node, which if we are to follow the original intent of neural nets it is to identify and explain the impact of inputs on the hidden nodes, since they are the only direct impact on the output nodes.

Therefore unless we understand the not so obvious hidden motivations of individuals and markets, we can never surely predict an outcome.

I like the idea that the adaptive NNs are capable of learning and understanding, and therefore making a choice by a learned algorithm. Many try to manipulate this; but it doesn't take into consideration those darn pesky hidden nodes which drive neuropsychology in all aspects. I am cautiously optimistic in believing it may drive our economy too; but I just need to figure out exactly how. Large data sets are needed, and parallel implementations are a must - but this requires an imperfect market as opposed to a tangible perfect one; which economists don't seem to enjoy very much.

Nonetheless, I think this is interesting stuff indeed.... maybe we need Recurrent Networks with bidirectional data flow to really understand this market...and right now we have unidirectional inputs without any transparency as to the hidden node.

Fun stuff.

Dani