Are there any guidelines for making comprehensive predictions?
Calibration is good, as is accuracy. But if you never even thought to predict something important, it doesn’t matter if you have perfect calibration and accuracy. For example, Google recently decided to restructure, and I never saw this coming.
I can think of a few things. One is to use a prediction service like PredictionBook that aggregates predictions from many people. I never would have considered half the predictions on the site. Another is to get in the habit of recognizing when you don’t think something will change and questioning that. E.g., I never would have thought not wearing socks would become stylish, but it seems to have caught on at least among some people.
Questioning literally everything you can think of might work, but it seems pretty inefficient. I’m interested in predictions which are important in some sense.
Yes, that’s one way to put it. The main restriction would be to pick “important” predictions, whatever “important” means here.
One other idea I just had would be to make a list of general questions you can ask about anything along with a list of categories to apply these questions to.
The main restriction would be to pick “important” predictions, whatever “important” means here.
I don’t know if there is any useful algorithm here. The space of possibilities is vast, black swans lurk at the outskirts, and Murphy is alive and well :-/
You can try doing something like this:
List the important (to you) events or outcomes in some near future
List everything that could potentially affect these events or outcomes.
and you get your universe of “events of interest” to assign probabilities to.
I doubt this will be a useful exercise in practice, though.
Are there any guidelines for making comprehensive predictions?
Calibration is good, as is accuracy. But if you never even thought to predict something important, it doesn’t matter if you have perfect calibration and accuracy. For example, Google recently decided to restructure, and I never saw this coming.
I can think of a few things. One is to use a prediction service like PredictionBook that aggregates predictions from many people. I never would have considered half the predictions on the site. Another is to get in the habit of recognizing when you don’t think something will change and questioning that. E.g., I never would have thought not wearing socks would become stylish, but it seems to have caught on at least among some people.
Questioning literally everything you can think of might work, but it seems pretty inefficient. I’m interested in predictions which are important in some sense.
Any ideas would be appreciated.
Are you asking how to generate a universe of possible outcomes to consider, basically?
Yes, that’s one way to put it. The main restriction would be to pick “important” predictions, whatever “important” means here.
One other idea I just had would be to make a list of general questions you can ask about anything along with a list of categories to apply these questions to.
I don’t know if there is any useful algorithm here. The space of possibilities is vast, black swans lurk at the outskirts, and Murphy is alive and well :-/
You can try doing something like this:
List the important (to you) events or outcomes in some near future
List everything that could potentially affect these events or outcomes.
and you get your universe of “events of interest” to assign probabilities to.
I doubt this will be a useful exercise in practice, though.
Yes, upon reflection, it seems that something along these lines is probably the best I can do, and you’re right that it probably will not be useful.
I’ll give it a try and evaluate whether I’d want to try again.