Hacker News Reader

Comment by thethirdone | Hacker News Reader

> The simplest explanation that best fits existing data, and makes useful predictions for the future.

My main issue with that definition is what makes it a useful prediction. Does a theory (for particle physics) that predicts the Higgs Boson more useful than one that didn't? If not, then before we collected the data to reject other theories they all would have been equal.

I would rather suggest that the properties of a good theory are best fitting existing data and having specific predictions for future measurements.

> Falsifiability, in this perspective, isn't a requirement in and of itself - it's a natural consequence of the requirement that theories be useful.

If you are of the opinion that any good theory is falsifiable (in the sense described in the article), then presumably you disagree with the rejection of falsifiability in the article. Where would you say the error in logic in the article is?

View on HN

Here's one significant error:

> Suppose that, contrary to what actually happened, Cavendish’s experiment had yielded a result that was qualitatively incompatible with the law of universal gravitation. The reaction of physicists would certainly not have been to reject Newton’s theory of gravitation and start looking for another one.

Looking for another theory is precisely what physicists do, once experimental error, miscalculation, and misunderstanding the theory are ruled out. Hence relativity and quantum physics, and the current search for a theory that explains dark matter and dark energy.

And another:

> From a purely logical point of view, what scientists do when they blame the failure of a prediction on the background assumptions instead of theory they take to be well-established is no different from what charlatans peddling pseudoscience do...

This is arguing that "falsificationism" requires discarding a theory after a single failed experiment, but forgets that even experiments can be falsified (by finding an error in the experiment or the associated conclusions).

The difference between scientists and "charlatans peddling pseudoscience" is that when a prediction proves to be false, scientists look for an explanation.

The author forgets that the explanation might be found in the experiment or the prediction, and scientists look there first before questioning a theory that's proven itself in the past.

A useful prediction is a prediction that matches experimental results.

You can't possibly know what will match experimental results in the future so that doesn't help determine good theories in the present.

And I would say that Newtonian gravity was a useful theory despite not matching experimental currently because when it was the best it mostly explained observations.

> Newtonian gravity was a useful theory despite not matching experiment

"Useful" in this context means useful for a future theory or experimentation, not utility. Newtonian theory certainly predicted many results, and all of them were useful in the above sense. A prediction doesn't have to exactly match observations to be useful, but that the theory produces a prediction which _could_ be tested against.

Even flat-earth theory makes useful predictions - albeit already proven false as time and time again, it predicts the wrong results.

What isn't useful is an unfalsifiable theory, which means the predictions it makes is not able to be tested, or the results of such a test could be construed to match if you squint a bit. A theory like creationism, for example.

I am a bit confused by your comment. I'm not sure if you agree with gliop or not. I am simply saying that Newtonian gravity is useful (in pretty much any sense) and that it did not match all experiments. Particularly in the orbit of Mercury could not be explained.

> Even flat-earth theory makes useful predictions - albeit already proven false as time and time again, it predicts the wrong results.

I definitely agree usefulness is independent of fitting experimental data. gliop would seem to say flat earth theories are not useful.

I don't think the value of a theory can be evaluated in isolation. It can only be evaluated relative to its competition. During its time, Newtonian gravity, even if it didn't perfectly fit all data, was still superior to its competition on an aggregate of the 3 criteria mentioned. Only once Einstein's theories were put forward, were we able to replace Newtonian gravity with a superior alternative.

Reply on news.ycombinator.com