The story of thinking and judgment in brain science is connected to behavioral science .
In decision theory , rational and perfect thinking is sometimes explained, but it is not always the best. The process on this page is used especially in situations that require momentary thinking.
The effect of making decisions based on previous experiences
Even if there is a rational and optimal response plan when it is necessary to respond, it is meaningless if it takes time to think about the response plan and the response cannot be completed in time. Such situations also occur in everyday life.
Human decisions are often based on experience rather than rational conclusions. The importance of heuristics is speed.
The mechanism of heuristics is studied in environmental psychology , cognitive psychology, psychology of decision- making, and behavioral economics .
The method known as machine learning seems to be a heuristic method because it learns as a fact if it is statistically more, regardless of whether it is rational or not.
For example, if there is a lot of data such as
"There was an egg in breakfast" and "It will rain in the afternoon "
, the rule of thumb is
"Egg in breakfast --> rain in the afternoon" .
For example, when conducting technical research in the field of quality studies , it can be dangerous to draw conclusions or assert anything based solely on heuristics. Considering heuristics as hints and backing them up with principles and the like can lead to innovative and robust investigations.
If we apply the mechanism of human cognition and learning and thinking and judgment to artificial intelligence (AI) , it looks like the diagram below.
Even methods such as regression analysis , which were originally devised in areas other than research on human mechanisms, can be applied to artificial intelligence. For example, even though Y and X can only be considered unrelated, if they are highly correlated , using this as a heuristic is a usage that is close to how humans work.
However, in the field of causal inference , it is often said that "correlation and causation are different." It is necessary to pay attention to the point "as a heuristic" to the last.
As with analogy commentaries, there are stories in phenomenology in which humans learn from analogies.
Even if they are in completely different fields, humans have the skill of complementing one another when the schemas are similar. Because you can do this, you can start from scratch, even in completely different fields.
Reinforcement learning and association analysis can be used on data collected in the same way as the data at which they were learned, or on that data, but it is difficult for things in completely different fields. .
Humans have a network structure and an ontology with a tree structure, and add (map) what they have learned to it, so humans can handle highly abstract schemas. maybe. One such method is the application of category theory.