welcome: please sign in
NeuralBasisOfAddiction

Other work (just discovered): Intermittent reinforcement (B. F. Skinner). A more general article comparing the different kinds of reinforcement is here: Reinforcement (Wikipedia)

Concept

I think that addiction could be the result of a glitch in the way our brain works.

Basically, the mind works by seeking pleasure and avoiding pain. There's a lot more too it, obviously, but that's basically how it works.

Our brain seems to stop us from doing one thing that we like continuously by making the pleasure diminish over time (or by having other desires override the current one).

Which is why addiction is a strange thing. When we're addicted to something, we end up repetatively doing something that is, on average, painful to us. It only sporadically gives us pleasure, and these sporadic payoffs make us want to work even harder and endure even more pain to get them.

A great experiment that illustrates addiction was done with mice. The mice were given a button that they could press to release tasty mouse food. However, the button didn't always work, and this was the experiment part; they were testing different random payoff probabilities to see what the mice would do.

When the thing paid off too often, the mice would fill up, and get bored of it and only press it when they were hungry. When it paid off too seldom, they wouldn't see the point in pressing it and give up.

However, if you gave them a small amount of payoff (about 30%), they became addicted to pressing the button! They'd also develop strange, superstitious, ritual behaviours in an attempt to figure out what was causing the machine to pay off.

It's very similar to what humans do in a casino at slot machines.

My theory is that if there's a small amount of payoff, your brain gets tricked. See, the brain is designed to descriminate and learn a lot of variables simultaneously. When you're learning to throw a ball at a target, there are many factors that could be affecting your throw:

When learning a task with many variables, you have to figure out how to isolate and analyze each one individually while not knowing how big an impact the other ones are having. For example, if your perception of distance is off, you won't even know if you're throwing it properly.

Of course, it's all quite complicated, and I admit that I don't totally understand how my brain learns, probably because of the recursive nature of the problem, but I've noticed some rough patterns.

However, I do believe that given what little information I have, I can draw a believable conclusion -- that the reason mice are becoming addicted to the food-button is because of this learning process.

I imagine that every time the mouse pushes the button and gets food, it's thinking "wait, why did I get food that time?" Then it's analyzing events in its environment to see what might've changed that made the button release the food (this is happening neurologically, of course -- I doubt mice are capable of that kind of conscious thought).

So, the mouse is correllating events in its environment with the food being released. The mouse is thinking, "maybe I pressed the button harder, or maybe my head was at a different angle, or maybe it was that the light got slightly dimmer for a second, or maybe it was that I did 3 quick presses followed by a slow press.."

Since the mouse isn't getting too much food (which causes the pleasure response to attenuate (i.e. it gets sick of the food)), the pleasure response is strong whenever it gets the food, so it really really wants to get it. And since the mouse doesn't have much brainpower, it can't quite grasp all of the possible correllations in its environment, and probably thinks that it's missing something. So it keeps on trying and trying!

The way I figure it, it's creating a new neural context in its brain for every success that doesn't match a known pattern. The brain works by strengthening patterns, and if the pattern doesn't exist, it forms a new one. So, it keeps making more and more patterns, and eventually it runs out of space, so it has to start overwriting patterns; I think that's where the "I'm probably missing something" feeling comes from.

What's scary is how similar this is to gambling in humans. People sit at slot machines trying over and over to win. They all have little supersitions and techniques that they try to make the slot machines pay off -- for example, always using the same machine, or wearing a lucky charm, or jumping from machine to machine. They think that there's a pattern to the randomness and they keep on trying to figure it out.

Information theory explanation

When doing statistical reasoning, an interesting problem called 'explaining away' can occur. The problem occurs when you try to reasoning about events that have two possible causes, but you can't know which occurred from the event alone. A common trap for machine learning algorithms is to "explain away" the event as being the result of the most probable cause. Another problem is when an event can be the result of its two causes simultaneously.

It's reasonable to assume that during our brain's evolution, it had to adapt a learning algorithm to tackle the same problem. The 30%-addiction-principle could be a side-effect of the heuristic that our neurons use to determine whether an event was caused by randomness, or a combination of weakly associated causes. (This would also explain the superstitious behaviour that people exhibit when stuck in this loop.)

Comments?

NeuralBasisOfAddiction (last edited 2010-04-24 09:29:36 by localhost)