4 ms·
To the people saying that correlation does not imply causation: we're Bayesian machines, not publishing journals. Feel free to weight this by whatever odds you
by radu_floricica 6y ago
To the people saying that correlation does not imply causation: we're Bayesian machines, not publishing journals. Feel free to weight this by whatever odds you give to causation being the other way, and act on it. For example if you think it's 2:1 as likely to be something else, like a confounding variable, it still a 10% decrease in all-time mortality to start eating chili. I'm very much averse to spiciness, but guess who's going to be considering it from now on?
- bookofjoe 6y agoI like it. From now on I'm going to refer to myself as a Bayesian machine.
- abecedarius 6y agoThat'd depend on your prior, before hearing of this, for expecting a mechanism for chilis to reduce mortality by 25%. My prior if you'd asked me would've been "c'mon, you really think so?" Not impossible but quite unlikely.
- radu_floricica 6y agoOf course. And a real-life bayesian agent should also have a shortcut for "odds too low to bother updating". I just wanted to point out that journal standards shouldn't be applied for real life decisions. What priors you have is totally up to you.
- abecedarius 6y agoI'm just saying your 2:1 for-example seems really optimistic. That'd be easy to miss.