Hacker News Reader

Comment by zak_mc_kracken | Hacker News Reader

I'm a big Sir Arthur Conan Doyle fan, but I'm always bothered whenever this quote surfaces:

> Once you have eliminated the impossible,” the fictional detective Sherlock Holmes famously opined, “whatever remains, however improbable, must be the truth.

This is terrible advice, even more so when mentioned in a scientific publication. This recommendation is the antithesis of what the scientific method is about and a text book example of an argument from ignorance.

You don't pick an explanation because you have run out of ideas to explain a phenomenon, for the simple reason that there might be other explanations you have not considered.

The time to believe something is when there is evidence to back it up.

View on HN

The key word in that quote is "eliminated". Try this more modern translation:

"Once you have falsified every other possible hypothesis, whatever hypothesis remains is the one we'll promote to a theory."

Holmes' is not saying you get to just make things up - he's trying to explain that instead of trying to prove something directly, inverting your perspective and falsifying as many incorrect interpretations ("the impossible") should always work - even if yo don't understand why.

That quote is an incredibly efficient summarization of the scientific method's core philosophy that (because you can't prove a positive), you falsify wrong hypothesis instead.

> That quote is an incredibly efficient summarization of the scientific method's core philosophy that (because you can't prove a positive), you falsify wrong hypothesis instead.

I disagree and I'll repeat what I said above: this is the antithesis of the scientific method. It's faith based and it opens the door to accepting hypotheses that have zero evidence to back them up. Follow this reasoning and you can just make things up.

In what way is that faith based? It seems to me that falsifying hypothesis until only one remains seems to be pretty scientific. Especially considering that (with regards to science, not solving a murder mystery) that means the world is forever stuck with whatever hypothesis was left at the end because we just didn't think of something else... the same principle or technique can always be used again...

The problem is that, in practice, there is no way to guarantee that you have actually eliminated every possibility. There are ways to constrain the answer within physics, but that doesn't always mean you have an explanation. Besides, within the scientific method (at least in my field of physics) emphasis is placed more on the predictions the answer can make, rather than the simple explanations. Simply eliminating other answers might give you an explanation for a singular case, but it fails to give you the predictive power we demand from all theories.

You gotta admit it works with mathematical proofs by contradiction.

True, but rarely do mathematical proofs translate into physical proofs.

Not really. Proof by contradiction means starting with the negation of the hypothesis you're trying to prove and then reach an impossible conclusion, thereby proving your hypothesis was wrong.

It still involves a rigorous logical reasoning, not accepting something because you're running out of ideas.

If you're defining the technique as not using logic, merely running out of ideas, of course its not logical and wouldn't work in practice. I don't see how what you described as not being 'eliminating the impossible', personally.

> emphasis is placed more on the predictions the answer can make

This is still based on falsifying the hypothesis; prediction is one of the useful tools by which a hypothesis can be experimentally tested. If a prediction ends up being incorrect, you're left with the null-hypothesis. It's entirely possible that the null-hypothesis may not be particularly useful, but it - as Holms said - "must be the truth".

At it's core, the scientific method is simply proposing ideas, and testing them. As zombie-Feynman said. "everything else is bookkeeping"[1].

[1] http://xkcd.com/397/

Just because you cannot reject the null hypothesis after an experiment (or 1000), does not mean it is true.

>there is no way to guarantee that you >have actually eliminated every possibility

So, what? You just stop once you eliminate the known possibilities?

That last possibility REQUIRES experiment, investigation and thought to verify. If you discover a new possibility then you apply the process again.

The approach is not a one-shot deal, you MUST keep applying it. The current laws of physics are models that we use to explain what we observe and (if the model is good) predict what we might observe.

This by no means implies that that the models describe what's really going on. It is possible to have a model that describes a system that works but is not complete. The deeper we dig, the more we learn.

The fictional quote was meant to describe a deductive approach to investigation, not a rigid method. How often did Holmes dig up new facts and change his mind after eliminating possibilities?

In the context of murder investigations, perhaps it is easier to form a complete set of possible explanations before applying that principle? That's how I always read it at least.

That's not how it's being interpreted here though. Everything that could possibly happen in the universe does happen, which means that anything which does not happen must be impossible. So we should go find all the rules that make things impossible, and we will be able to predict what will happen from what's left.

Indeed. Plus, there's another subtle moment in science. It just creates models which are good at matching observations and making predictions.

Many mutually conflicting models can be created to describe the same properties of the same system.

We have no way of ever finding the "truth".

That implies that there is a priviledged truth. If two theories are completely indistinguishable by any conceivable experiment, then I would consider them both to be equally "true".

How would they not be the same if they are indistinguishable by any experiment?

By using completely different mathematics.

Would it be reasonable to claim that if different mathematics give the same answer in all cases, then it is really only different forms of a single mathematic?

They could produce different answers for cases that don't occur in nature. For example you can find many polynomials that go through a given point set (all possible observations you can make) but disagree wildly outside of the point set.

Define "completely different" please. The two mathematical representation of quantum mechanic were completely different until it was shown that they were equivalent..

Yes that's what it implies. There's a privileged truth and we don't get to know it. Don't feel bad, though.

You can describe the laws of physics like little gremlins who follow a book of laws that happens to be the laws we know, and the predictions and observations we have will match the gremlins theory.

It doesn't make the existence of tiny gremlins real.

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

Reply on news.ycombinator.com