Thursday 22 June 2017

Operant Conditioning

I don't know how many of you are familiar with real world psychological theories but there's one theory called operant conditioning that I was musing over the other day (actually I had to learn about it for school but whatever) and I realised with that it actually fits the Creatures model of learning surprisingly well.

Here is a fairly decent article about the theory if any of you would like to read about it in detail. While I will be summarising the main points here, I figured a more detailed explanation might be of interest to some of you.

The theory of operant conditioning was first researched by American psychologist B. F. Skinner who performed various experiments on real life pigeons. His experiments attempted to teach the animals to carry out certain (voluntary) behaviours in response to particular stimuli. And he found that this was easily achievable by manipulating the consequences of the behaviours: if performing the action had a good consequence (such as food) the pigeons were more likely to repeat the behaviour. This is the essential basis of operant conditioning: someone (whether they be a human or an animal) learns to perform a certain behaviour in a certain situation because when they tried that same behaviour in that same (or a similar) situation before, it resulted in a favourable outcome. Operant conditioning can also work to dissuade a particular behaviour in a particular situation if the consequence is an unfavourable one.

I've written about the Norn learning process before so you can look at that article if you want to see the process more in depth than a brief summary. But how Norn learning works, essentially, is that there's a lobe (the concept lobe) in which the neurons memorise certain situations, such as "There is food nearby + I am hungry". Through the dendrites (neural connections) from the next lobe (the decision lobe) the Norn brain links certain situations to certain actions and in this link is encoded the information of whether or not the Norn should attempt or avoid that action in that situation. These links are strengthened by the reward and punishment chemicals, which are released to signal a favourable outcome from an action (such as reducing hunger) or an unfavourable one (such as experiencing pain). The stronger the link between the situation and the action the longer the connection will stick around and the more likely the Norn will be to perform (or avoid) that action in that situation.

Sound familiar? I could have sworn I just explained an almost identical process a few moments ago... look at this handy summary table of the steps involved in each learning process:

Operant conditioning
  1. A particular situation is in place
  2. An action is taken in that situation
  3. The consequence of that action (favourable/unfavourable) affects the likelihood for the action to be repeated the next time that situation is encountered
Creatures
  1. A particular situation is in place
  2. An action is taken in that situation
  3. The consequence of that action (favourable/unfavourable) affects the likelihood for the action to be repeated the next time that situation is encountered

I rest my case. I don't know if Steve Grand had operant conditioning in mind when he designed the Norns; it wouldn't surprise me if he did, considering how accurate it is to the theory. Either way it just goes to show how lifelike Norns are.

No comments:

Post a Comment