Hacker News Reader

Comment by Cats_In_Love | Hacker News Reader

Probability is fundamentally all we have in science. The only distinction is in the level of confidence use in inference of the data. 95% is fairly typical though particle physics requires and impressive level >99.9999%

View on HN

even if one accepts this, which I don’t (what is the probability of mathematics being a quantum ultrafinitist glandular endomorphism of classical electromagnetism? What is the probability that the sun rises tomorrow? What’s the probability the standard model is true? It’s nonsense.) you could totally interpret what he said as “original paper claimed 20-80% confidence, economist article assumed 95%” and that would be a reasonable probabilistic reading

One way of considering the things you describe in terms of probability is to frame them as bets, and then set the probabilities based on what bets you might make (e.g. what odds would you want in a bet on whether the sun would rise tomorrow). Personally I find that makes some of the more difficult statements a little more palatable.

https://en.wikipedia.org/wiki/Bayesian_epistemology

the point being that IMO the way you understand and come up with and reason about something like “classical mechanics” or “timeless decision theory” or “quantum enlightenment time cube satanic world order simulation” or “risc v architecture” is not about probability, and saying science is about probability totally elides the actual complex information and reasons that go into creating and using those in favor of just saying “they have a probability and we can change them”.

> what odds would you want in a bet on whether the sun would rise tomorrow

How would you denominate such a bet? I don't think money would have much value in a world where the sun failed to come up one day.

I understand what you’re saying, but I have to highlight a mistake.

In science, what you’re talking about is not a “probability of being right”, but a probability of not getting the same experimental results completely randomly, without any underlying cause. You still might have 0% probability of being right. With “95% confidence” there still could be no measured effect whatsoever, you just made the same experiment multiple times and randomly finally got big enough random numbers to get you 95% confidence.

It’s not a nitpick, it is a serious mistake that 95% social scientists make.

Reply on news.ycombinator.com