The Problem of Induction (1967)

One way is to eliminate all other possible options. I'm not quite sure that the evolution example is a good one. We have the opportunity to observe some parts of evolution.

All knowledge is contingent; when the facts as I understand them change, I change my opinions!

It seems like he is writing more about "Correlation doesn't imply causality" in the black balls example. And that we can't generalize just from trends. This is similar to Taleb's Black Swan, or the turkey's weight trending upwards until Thanksgiving.

In terms of Evolution, the theory does give us predictions, and we are able to run experiments to show that it does happen.

I must also comment that a screen name of WhoBeI is very appropriate for a discussion on philosophy of knowledge. :-)

>philosophy of science in its long history produced anything that has proven useful in the practice of science

The scientific method comes to mind.

Ernst Mach has certainly had real significant influence on the physicists of his generation and the physics they did:

http://en.wikipedia.org/wiki/Ernst_Mach

Yes, absolutely. The modern definition of "proof" and of "true". Logic was considered philosophy in the early 20th century. And the works of Russell, for example, helped define what we see today as a convincing proof. Basically, 120 years ago, math was not as rigorous as it is now (Cauchy is a prime example).

Also, philosophy of science is not necessarily about producing useful stuff for the practice of science, but usually helps better understand the relationship between humans and knowledge. It allows scientists to take a step back and see the big picture before diving back into their practice. As such, it shapes the way we think of and do science, and allows for better questions to be asked, and gives better ways to frame those questions.

A better understanding of what it is they are doing. I am currently reading "The Structure of Scientific Revolutions" (http://en.wikipedia.org/wiki/The_Structure_of_Scientific_Rev...), and I have difficulty imagining how it could not inform a scientist-reader about their own methods. (Except if they had already encountered these ideas distilled elsewhere.)

It's really hard to talk about "truth" anywhere, but it's really crucial to both talk about and agree upon the meaning of "truth" in science since, ultimately, that's the goal of the scientific process—the pursuit of the thing.

Karl Popper was a major study of this problem and wrote about probably the most common and universally acceptable notion of scientific truth in the mid 1900s. Before then this notion of truth was in practice used, but it may have been difficult to have a clear conversation about it. Worse, without a clear understanding of the properties which give rise to processes with generate "truth" (i.e. the "scientific method" which, even being really generous, is better thought of as a menagerie of related methods) it may have been a little hard to determine whether someone's scientific program would be effective or agreeable.

So Popper clarified the notion of "Science as Falsification", of "Truth as what weathers evidence". For Popper, nothing was science unless it was vulnerable to falsification and nothing was true unless it was both science and withstood all attempts to falsify it.

This has a major effect on modern statistics. Statistics can be regarded, not entirely incorrectly, as the quantification of the pursuit of truth. It seems to give mechanism to the idea of models relating to evidence and the pruning out of falsified models. It's difficult to ask questions about why statistics is a meaningful tool for scientific pursuits without invoking Popper.

One important sociological constant for a couple centuries is if a hypothesis comes from a daydream, like the ring resonance structure of benzene, and its proven right, it comes from science. If a hypothesis comes from a philosophical argument, like the atomic theory of matter, and its proven right, we very carefully forget about where it came from and don't discuss it, very "Fight Club" like.

So its not like nothing has ever come from philosophy as a hypothesis successfully validated as a useful predictive model, its that it happens all the time and we do NOT talk about it, just like Fight Club.

(edit to add another way to phrase it is if a hypothesis is dredged out of one individuals unconscious random internal noise source, we call its source Science. If it comes from a rational and reasoned conscious debate between thinkers, we don't talk about it at all, or at best we call it common knowledge. Perhaps its an attempt not to appear vain or conceited?)

Here's an example of where the philosophy helps. You would probably agree that physics is a clear science and that its methods—philosophically reasoned about—produce (or discover) genuine truth. But would you agree that a "softer" science—say, psychology—still has enough methodological rigor to produce genuine truth? How about sociology? Economics? The more observationally ambiguous—the more abstract, profoundly complex, or "fuzzy" a science's subject is—the less confident we are that it's really a science, that it really produces truth. So how do you reason about what sciences or individual experiments are productive and which are flawed, uncertain or, frankly, bogus? Philosophy!