@jodiecongirl It makes sense to have drivers precommit to being oncall to work a disaster a year to keep their Uber job.
A market-clearing price at surge prices may bring a few more drivers but only to the richest customers.
Biological psychiatry, a total failure. Thomas Insel, the director of the US National Institute of Mental Health from 2002 to 2015, said in an interview in 2017: “I spent 13 years at NIMH really pushing on the neuroscience and genetics of mental disorders, and when I look back on that I realize that while I think I succeeded at getting lots of really cool papers published by cool scientists at fairly large costs - I think $20 billion - I don’t think we moved the needle in reducing suicide, reducing hospitalizations, improving recovery for the tens of millions of people who have mental illness. I hold myself accountable for that.” https://t.co/YpqYhDZNLb
Paul Meehl talked about putting theories at "grave risk of refutation" through prediction. Deaton & Cartwright similarly ask we wager a prediction to falsify. A good experiment should have done all the preparation to kill your theory. But if not, make a prediction to kill it.
A frequentist bootstrap in Bayesian terms means the data are exchangeable. No value outside the sample is possible. The prior adds no information. Rubin (1981) asked whether anyone would choose that prior on purpose. Often the answer is 'no'. Why not use Bayesian Bootstrap?
@LongFormMath I haven't read all the answers but the formula
P(win) = P(yes)P(win | yes) + P(no)P(win | no)
governs both cases (law of total probability)
Red: P(win) = ½ (1/26) + ½ (1/26) = 1/26
Ace Spades: 1/52(1) + 51/52(1/51) = 1/26
The dominant model people use to think about AI's effect on the labor market is automation. But perhaps commoditization is a better model. Automation implies job apocalypse. Commoditization implies plentiful jobs with low wages. 1/
Aggregated knowledge from a small number of debates outperforms the wisdom of large crowds
This study asked a live crowd (N = 5,180) to respond to general-knowledge questions (eg "What is the height of the Eiffel Tower?")
Participants first answered individually, then deliberated and made consensus decisions in groups of five, and finally provided revised individual estimates.
Averaging consensus decisions was substantially more accurate than aggregating the initial independent opinions. In short, small groups beat the wisdom of crowds.
Combining as few as four consensus choices in small groups outperformed the wisdom of over 5,000 individuals.
The key was reasoning together--that aspect of collective deliberation allowed small groups to make more accurate estimates than massive crowds.
https://t.co/Yz23ELwgde
Amazing new paper by Chetty and the Opp Insights team — digital twins can recover causal effects
They validate the digital twins (ie rich ML applied to generate counterfactuals from lots of covariates) with RCTs, allowing for scalable program evaluation
https://t.co/9QCEEKiiWA
@johnjhorton If m*U(airheads, sour patch kids, etc) = n*U(chocolate), where m > n, then he owes you (m/n)*chocolate tax units in airheads, sour patch kids, etc..