The thing is, some of the steps are very vague
You’re right of course, this was meant to be fully general. Details should be tuned on each specific instance.
If you have a bad case of insufficient clue, what’s the cure?
I’m not sure I understood what you mean, but I guess you’re thinking about cases where you can’t have a “perfect experimental setup” to collect information. Well, in this case one should do the best with the information one has (though information can also be collected from other external sources of course). Sometimes there’s simply not enough information to identify with sufficient certainty the best course of action, so you have to go with your best guess (after a risk/reward evaluation, if you want).
Well, it’s somewhat hidden in steps 2 and 3. You have to be able to correctly state your hypothesis and to indentify all the possible variables. Consider chocolate water: your hipothesis is “There exist some brands of water that tastes like chocolate candy”. Let’s say for whatever reson you start with a prior probability p for this hypothesis. You then try some brands, find that none tastes like chocolate candy, and should therefore apply bayes and emerge with a lower posterior.
What’s much more effective, though, is evaluating the evidence you already have that induced you to believe the original hypothesis. What made you think that water could taste like chocolate? A friend told you? Did it appear in the news? In the more concrete cases:
Sex partners : Why did you expect them to be able to satisfy you without your input? What is your source? Porn movies?
Computer repair shops : Why did you expect people to work for free?
Diets : Have you talked to a professional? Gathered massive anedoctale evidence?
This can be easily generalized as an algorithm.
Something repeatedly goes wrong
Identify correctly your prior hypothesis
Identify the variables involved
Check/change the variables
Observe the result (apply bayes when needed)
Repeat if necessary
Scientific method applied to everiday life, if you want :)
The thing is, some of the steps are very vague. If you have a bad case of insufficient clue, what’s the cure?
I’m not sure I understood what you mean, but I guess you’re thinking about cases where you can’t have a “perfect experimental setup” to collect information. Well, in this case one should do the best with the information one has (though information can also be collected from other external sources of course). Sometimes there’s simply not enough information to identify with sufficient certainty the best course of action, so you have to go with your best guess (after a risk/reward evaluation, if you want).
Sorry, I wasn’t very clear.
I meant that if you have a deep misunderstanding of what’s going on, as here, what do you do about it?
Well, it’s somewhat hidden in steps 2 and 3. You have to be able to correctly state your hypothesis and to indentify all the possible variables. Consider chocolate water: your hipothesis is “There exist some brands of water that tastes like chocolate candy”. Let’s say for whatever reson you start with a prior probability p for this hypothesis. You then try some brands, find that none tastes like chocolate candy, and should therefore apply bayes and emerge with a lower posterior. What’s much more effective, though, is evaluating the evidence you already have that induced you to believe the original hypothesis. What made you think that water could taste like chocolate? A friend told you? Did it appear in the news? In the more concrete cases:
Sex partners : Why did you expect them to be able to satisfy you without your input? What is your source? Porn movies?
Computer repair shops : Why did you expect people to work for free?
Diets : Have you talked to a professional? Gathered massive anedoctale evidence?