Two Choice Task

A two choice task is given an input choose one of two possible actions.
The answer can be learned via reinforcment.
This can be probabilistic
Prior work, from Friedman et al. in 1964, showed that people roughly follow the probabilities but lean toward 50%.
We used reinforcement learning with environmental feedback, and some randomness to reproduce this work.
The trick was to punish incorrect decisions.
If you don't do that, just by Hebbian learning, the incorrect decsion becomes more likely.
We did this by randomly activating the other answer.