IMHO LW grossly overstates the importance of statistical thinking.
If you have a job making decisions for groups then statistics matters greatly but in our personal lives, genuinely statistical problems are so rare compared to the huge number of decisions we actually face.
Most of our decisions are characterised by uncertainty, uniqueness and an absence of data. Real-world rationality is less like mathematics and more like system design where we weave together plans and strategies. For example using contingency planning - a rational buyer ensures the seller has a good return policy rather than trying to use Baye's rule to analyse their past purchase experiences.
IMHO the overlooked core of rationality is creativity - you have to imagine possibilities and invent plans. Its only when you reduce decision-making to math problems that math seems so important.
Personally I always understood LW's focus on statistical thinking as something similar to what math classes do, from times table to solving integrals - what matters is that a) you have the understanding of underlying concepts which makes you think in a particular way, and b) training / learning caches some operations in your head which lets you use new models of thinking that you otherwise would not be able to (e.g. memorizing times table serves just to make most calculations feasible to do in head or on paper in reasonable time). You don't have to do a perfect Bayesian updates in your daily life - just understanding how the thing works + little training lets you not fall into obvious traps when learning new evidence.
You are exhibiting a case of failing in the "2. Ability to model other people as having really different mind-designs." The author is a psychologist, not a statistician. When he talks about thinking probabilistically, he is not talking about sitting down and doing the math. He is talking about recognizing that there is no such thing as certainty and understanding that even very unlikely outcomes may occasionally occur. It is about making reasonable rational decisions in an uncertain world, hedging uncertainty if possible, and generally being OK with it.
In college I had a friend who didn't understand 50/50. If an outcome was not yet decided, then both options are the same "size", right?
One would like to think that my debates with him about this, or maybe the stats class he took, we're what helped fix this intuition. But what probably was just as important was playing games like LoL, where in the face of uncertainty, you make calculated risks all the time. At some point, it is impossible to get better without making intelligent risks.
Client: You styled my house 3 months ago it is still not sold! You said it would sell much more quickly!
Stylist: Your house was moved from the population of un-styled houses to the population of styled houses. This population has a different mean time-before-being-sold but it still has long tails, I'm sorry you are in that tail.
Btw: "a rational buyer ensures the seller has a good return policy rather than trying to use Baye's rule to analyse their past purchase experiences."
Sounds to me like one would first look for priors that confirm that good return policy, one does not pull trust out of thin air.
Agreed. Having seen other try the statistical approach to making many of life's decisions I have concluded that it's somewhere between futile and or worse self deception.
It might work things that happen over and over again like deciding which route to take to work. You can accumulate enough useful data to do that.
Many new and large decisions by their nature do not have that data and then you're stuck looking for proxies for it based on other's data (experiences) and that rarely applies to your circumstances.
An example of this would be moving to SV, you prob don't have your own data points so you're relying on data of others (friends, people on the internets) but it's unlikley that that will have a bearing on how you feel about it once your there.
Even when our models are weak and there's a lot of uncertainty, putting numbers on it can clarify your reasoning and eliminate a lot of potential errors. E.g. the Drake equation consists of five or six numbers none of which we know, but it still tells us a lot.
Doesn't your example contradict your claim? Putting numbers into the Drake equation doesn't really make it more useful than just thinking about it abstractly.
I wonder if the discussion of that article would have gone differently if it had been about (internally) reasoning about uncertain things as 51/49 choices and made it clear that it isn't really useful to communicate using the made up numbers (except maybe if the whole hypothetical case is explicitly communicated).
> Doesn't your example contradict your claim? Putting numbers into the Drake equation doesn't really make it more useful than just thinking about it abstractly.
No? The way you (or at least I) use it is to plug numbers in (I mean, what else can you do with it?). "I think this factor is somewhere between 1 and 10,000, and this one's about 1,000,000, so... huh. My model can't be right."
> I wonder if the discussion of that article would have gone differently if it had been about (internally) reasoning about uncertain things as 51/49 choices
In a 51/49 scenario there's no point spending a lot of time evaluating. But unfortunately we often naturally assume a scenario is 51/49 when it's not.
> made it clear that it isn't really useful to communicate using the made up numbers (except maybe if the whole hypothetical case is explicitly communicated).
A number calculated from made up numbers is probably more accurate than a gut feeling, and we communicate those all the time.
If you have a job making decisions for groups then statistics matters greatly but in our personal lives, genuinely statistical problems are so rare compared to the huge number of decisions we actually face.
Most of our decisions are characterised by uncertainty, uniqueness and an absence of data. Real-world rationality is less like mathematics and more like system design where we weave together plans and strategies. For example using contingency planning - a rational buyer ensures the seller has a good return policy rather than trying to use Baye's rule to analyse their past purchase experiences.
IMHO the overlooked core of rationality is creativity - you have to imagine possibilities and invent plans. Its only when you reduce decision-making to math problems that math seems so important.