Maps and models are useful but influence in both directions: Both users and those making them. They should help us relate to the world.
The truly competent know what they don't know. To be competent to that degree you need to be eager to learn, talke to the "lifers" who have such deep knowledge, and solicit feedback.
Go to the irreducible parts constituting the problem. Can use socratic questioning:
- Clarify thinking and explain origins of your idea
- Challenge assumptions
- Look for evidence
- Consider alternative perspectives
- Examine consequences and implications
- Question the original question
Use "five whys".
Allow you to make rational and considered ways to imagine possible future scenarios.
Competent thinking also covers the second (and higher) order of consequences.
Garrett Hardin, 1963: "You can never merely do one thing."
Hardin's "Slippery Slope Effect" (wedge) - that action A will lead to (always?) negative outcomes B, C, D... - can be rebutted with pragmatism, and real life examples showing that everything has limits.
- Bayesian thinking: Collect data and adjust probabilities based on what we learn --> conditional probability, consider conditions preceding events
- Fat-tailed curves: Displays more extreme cases than e.g. bell curves
- Asymmetries: Metaprobability - how probable are your estimates even? Very infrequently people hit the mark in their expectations
- Assume that what you are trying to prove is true/false. Show what else would have to be true.
- Consider what you want to avoid; see which options are left.
What is the simplest, plausible solution to your problem? Beware that this concept does not always hold true: Some things are complicated.
Do not attribute to malice that which is more easily explained by stupidity.
Karl Popper - science requires strong ideas to be proven falsifiable.
It's not enough to simply have all necessary conditions in place.
"The set of conditions necessary to become successful is a part of the the set that is sufficient to become successful. But the the sufficient set itself is far larger than the necessary set."
Easy to mix them up or get wrong results.
Extremes tend to lead to a "regression to the mean" over time.
Control groups best way to validate causation.