Decision-making has captivated human intelligence for many years. Humans have always wondered what makes us the most intelligent animal on this planet. The fact is that decision-making could be seen as directly correlated with intelligence. The better the decisions being made, either by a natural or artificial agent, the more likely we will perceive that agent as intelligent. Moreover, the level of impact that decisions have is directly or indirectly recognized by our societies. Roles in which decision-making is a primary responsibility are the most highly regarded in today's workforce. If we think of prestige and salary, for example, leadership roles rate higher than management and management rate higher than the rest of the labor force.
Being such an important field, it comes at no surprise that decision-making is studied under many different names. Economics, Neuroscience, Psychology, Operations Research, Adaptive Control, Statistics, Optimal Control Theory, and Reinforcement Learning are some of the prominent fields contributing to the understanding of decision-making. However, if we think deeper, most other fields are also concerned with optimal decision-making. They might not necessarily contribute directly to improving our understanding of how we take optimal decisions, but they do study decision-making apply to a specific trade. For instance, think journalism. This activity is not concerned with understanding how to take optimal decisions in general, but it is definitely interested in learning how to take optimal decisions in regards to preparing news and writing for newspapers. Under this token, we can see how fields that study decision-making are a generalization of other fields.
In the following lessons, we will explore decision-making in regards to Reinforcement Learning. As Reinforcement Learning is a descendant of Artificial Intelligence, in the remaining of this chapter we will briefly touch on Artificial Intelligence. Also, being a related field, we will look at some basics of probability and statistics. On the rest of this lesson, we will discuss decision-making when there is only one decision to make. This is perhaps the major difference between Reinforcement Learning and other related fields. Reinforcement Learning relaxes this constraint allowing the notion of sequential decision-making. This sense of interaction with an environment sets Reinforcement Learning apart. In later lessons, we will continue losing constraints and presenting more abstract topics related to Reinforcement Learning. After this lesson, we will explore deterministic and stochastic transitions, know and unknown environments, discrete and continuous states, discrete and continuous actions, observable and partially observable states, single and multiple agents, cooperative and adversarial agents, and finally, we will put everything in the perspective of human intelligence. I hope you enjoy this work.