4 ms·
Ultimately, truth-seeking is a personal thing and you should tune your methods to the machine that is you. tl;dr SOWH is good. Don't claim it for yourself. Slo
by renewiltord 6y ago
Ultimately, truth-seeking is a personal thing and you should tune your methods to the machine that is you.
tl;dr SOWH is good. Don't claim it for yourself. Sloppy Priors are good. Be careful of not adjusting on evidence. The bet trick is very good. Be careful about your utility function.
Personally, among the people I know the following doesn't happen:
> In such cases, the failure mode is that ‘Strong Opinions, Weakly Held’ turns into ‘Strong Opinions, Justified Loudly, Until Evidence Indicates Otherwise, At Which Point You Invoke It To Protect Your Ass.”
And that is mostly because it's really easy to claim you're operating in SOWH and really hard to determine if you actually are. So most truth-seekers apply the same techniques of epistemology to it as they do to other things that require auto-evaluation with no objective truth set: you trust weighted external input more than your own. Am I charismatic? Am I intelligent? Am I whiny? It is hard for me to say. It is far easier for me to find that out from people I can identify as trusted on the subject.
I remain convinced that operating in SOWH is empowerment (at a 80% certainty har-har). I believe that the Bet Trick and the Sloppy Prior Trick are both clever techniques to improve your search for truth as well.
The Bet Trick is very good. My personal danger is that I do not apply linear utility (i.e. an unlikely win makes me much happier than a string of losses). So if I predicted that the US would be open by August and I'm right that makes me way happier than if I predict repeatedly that median summer temperature is greater than median winter temperature in Saskatoon. I'm okay losing $200 repeatedly on the former for the win.
The Sloppy Prior also has the trick that it allows you to examine your reaction to evidence. Your Sloppy Posterior to the weakest of evidence must be different from your Sloppy Prior! If it isn't, your cognition is currently failing you. The only problem is that the sloppiness gives you room to avoid having different posteriors.
That last part I find very hard in almost-certain and almost-never situations: if a set of instrument measurements show that temperatures across the Earth are the same as they were 40 years ago, then it's highly likely that the measurements are broken, but it is not certain, so my posteriors for AGW given that evidence should drop. But they don't unless I am conscious of this.