Researchers have confronted us so many times of how humans get things wrong when it comes to decisions making ... "Humans misunderstand probability, pay attention to the wrong things and just mess up ... "
Some psychologists believe that using heuristics, rules of thumb, and other "shortcuts" often leads to better decisions than the models of “rational” decision-making developed by mathematicians and statisticians.
So ... what is the right way to take a decision?
Lets review quickly what is in between reinforcement learning and decision making ... "Reinforcement learning is an area of machine learning inspired by behaviorist psychology, concerned with how software agents take decisions/actions in an environment to maximize their reward".
A model based reinforcement learning is a framework in which the software agent tries to understand the world and create a model to represent it. Using this statistical model, the agent plans and make his decisions/actions ... statistically ... to maximize his rewords.
A model free reinforcement learning is a framework in which the software agent tries to learn ... "intuitively" ... a policy that maps directly between inputs and decisions/actions ... to maximize his rewords.
If one compares the model based method to the "analytical human thinking", while the model free can be considered as the "human instinct", here is our advise for decision making ...
Instincts are useful for "simple decisions" where the problem states space is small, finite, clear and well understood ... don't be tempted to trust your instincts in case of complex decisions ... let the machines take them for you.