I do a lot of science. I'm an odd kind of computer scientist - a lot of times I work on optical systems tied to computers. I have first hand knowledge that OFTEN facts are in dispute: how did you calibrate, what are your tolerances, sensitivity, replicates, etc. Not just my interpretations of these facts are questioned, but the means by which they were collected and limitations, hidden relationships, whether they are confounded or not. When you get in to cutting edge science, many times there are huge problems with "the facts". They can and are disputed. For good reason. That's where you ultimately have to use logic, modified with probability. Symbolic logic is just a special case of random process analysis where the variance is zero (you know for sure the values are true or false). But the principles are general.
Gee, with Plan A we're 92.5% sure with a 2% tolerance. Plan B is 89.5% but with a 6% tolerance. What should we do? Get more data. Then all these get mushed into a risk analysis, or cost benefit analysis, where all these terms are weighted by money, or some other factor. My interest in stocks involved neural network analysis. I couldn't find a good signal. Lots of facts, but hard to interpret. I used neural networks to predict football. I did about as well as the line. No better. A little worse, in fact. But still, that's not bad for some Joe with a 486 (this was a few years back).
No gold under that rock. Yet. Wouldn't rule it out, though. SO truth, to me, has become this mushy thing. You never really get there. And many things people think they know are simply wrong. |
