Luftmensch
Senior Member
- Joined
- Dec 13, 2017
- Messages
- 2,316
- Reaction score
- 3,186
Unqualified, ‘evidence’ is a kind of weak term in math. Evidence that something is true is like, you had a computer check that it’s true for the first 20 cases, but there’s still infinitely many more cases to check.
I read this post 20 times and 20 times only. It was true each time... so by induction....
Out of meandering interest... 'Evidence' is an important term in bayesian statistics/inference. It is used a lot. Often we are unable to make direct observations of variables of interest. These are called hidden variables. Instead we have to do the best we can with incomplete, indirect or noisy observations. These weaker observations are called 'evidence'.
Generally, the uncertainty in a model decreases as you gather evidence. The opposite of uncertainty is belief - as you gather more evidence, your belief in a hypothesis get stronger.
One of the cool outcomes from Bayesian statistics is that you can never know something definitively. Instead knowledge is represented as probability distributions. To know something definitively, you'd need to collect an infinite amount of evidence. If you did, your distribution might collapse into something that looked like a dirac delta function. But this is 'in the limit'... instead we just have to settle for being "pretty bloody sure"...