One reason for irrationality is:
Avoid considering the situation or context for decision-making. That requires hard thinking.
Choose the most general approach, i.e. it makes the least number of assumptions about the context for its application and works in most contexts.
The most general approach is likely to be the hardest or less efficient, as it does not leverage the particularities of the current context.
Choose working hard over hard thinking. Label any thinking as “overthinking”. Aim at zero thinking, for “efficiency”.
We may consider acquiring knowledge as an example:
The easiest way to learn is by reading and thinking.
You need to research yourself only what nobody else has shared; prefer formal methods (thinking).
Doing mathematics empirically may seem easier than formally, but it is often as misleading as numerology.
Even when empirical methods are required, proper foundations are necessary.
When you need to venture into the unknown empirically:
Underthinking seems to be too prevalent nowadays. Avoid mindlessly digging your own grave thinking only that there is no other thing to do or a reason to think differently. Sometimes it would seem that there is a fear of thinking (phronemophobia).5
Particularly successful examples are cryptocurrencies and JavaScript.
Knowledge is connected. Understanding differential calculus is hard without previous knowledge about arithmetic. When two pieces of knowledge cannot be reconciled with each other, there is something missing, and possibly something wrong. Hence reconciliation is very important in sciences, e.g. unification in physics, and Principia Mathematica.
Something true is not necessarily all the truth. Consider the tree blind men and the elephant. Reading will give you information, but you still need to apply thinking to transform it into knowledge. There is no practical difference in the result of being misinformed or being properly informed but applying that information harmfully.
Shots in the dark may be useful if failure can be identified and corrected. Avoid digging your own grave deeper with a persistent delusion of success.
“All models are wrong, some are useful.” However, when using models we must be aware of their limitations. Properly defined models usually have explicit assumptions or ranges for parameters that constrain the contexts where it is applicable. Thinking that a model works where it does not is delusional knowledge.
The harder approaches work when the easier approaches are not available. But the easier approaches should take precedence when possible.
People seeking for silver bullets (only one approach for every case) will choose the hardest approaches. Avoid that.
This was one of the many points of the last day. Less succinctly and with references to many other things.