Less Wrong is a community blog devoted to refining the art of human rationality. Please visit our About page for more information.

Sniffnoy comments on Making Beliefs Pay Rent (in Anticipated Experiences) - Less Wrong

110 Post author: Eliezer_Yudkowsky 28 July 2007 10:59PM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (246)

Sort By: Old

You are viewing a single comment's thread. Show more comments above.

Comment author: Sniffnoy 29 December 2009 01:22:20AM 4 points [-]

Because one of these allows you to make predictions, and the other doesn't. Saying "fire has a cause, and I'm going to call it 'phlogiston'!" doesn't tell you anything about fire, it's just a relabeling. Now, if you make enough observations, maybe you'll eventually conclude that "phlogiston is the absence of oxygen" (even though this isn't really correct), but at that point you can throw out the label "phlogiston". Contrariwise, if you say "oxidization causes fire", where "oxygen" is a previously known thing with known properties, then this allows you to actually make predictions about fire. E.g. the fact a candle in a sufficiently small closed space will go out before it melts, but not necessarily if there's a plant in there too. One pays rent, the other doesn't.

Comment author: DanielLC 29 December 2009 02:17:02AM 2 points [-]

You can make exactly the same predictions with phlogiston. If you burn coal next to iron, it will refine it. You could predict this with oxygen (oxygen is moving from the iron to the coal) or with phlogiston (phlogiston is moving from the coal to the iron).

It's like with electric charge. If you think of it as positive charge moving around, it has almost exactly the same predictive power as thinking of it as electrons moving around.

Comment author: Sniffnoy 29 December 2009 05:42:33AM 2 points [-]

But you can only predict it if you already know that a gain of phlogiston refines iron; if you don't, you can only observe it afterward and write it down as a property of phlogiston.

If you don't know anything about oxygen or phlogiston beforehand, then, sure, they're pretty much equally predictive, i.e., not very much. But if "oxygen" is not in fact just an arbitrary label as "phlogiston" is, but in fact something you're already working with in other ways, then they're not symmetric.

Also as Nick Tarleton points out below there are other asymmetries, though those are not so much in the predictive power.

Comment author: DanielLC 31 December 2009 12:42:43AM 0 points [-]

"But you can only predict it if you already know that a gain of phlogiston refines iron"

Same goes for oxygen.

Comment author: DanielLC 31 December 2009 01:47:12AM -1 points [-]

Okay, I admit that that's not really a prediction, but until then, they couldn't even explain it.

If you're going to do it like this, what's one thing oxygen predicted?

By the way, I'm responding to the fact that I lost two karma points on that, not any actual post.

Comment author: Sniffnoy 31 December 2009 01:59:17AM *  2 points [-]

That's what I just said.

Comment author: DanielLC 31 December 2009 02:25:24AM 2 points [-]

Sorry. Too used to defending my position to realize you're not attacking it.

Comment author: Nick_Tarleton 29 December 2009 02:43:28AM 10 points [-]

Because one of these allows you to make predictions, and the other doesn't. Saying "fire has a cause, and I'm going to call it 'phlogiston'!" doesn't tell you anything about fire, it's just a relabeling.

The hypothesis went a little deeper than that. "Flammable things contain a substance, and its release is fire" lets you make many predictions — e.g., that things will burn in vacuum, or that things burned in open air will always lose mass (this is how it was falsified).

Comment author: Sniffnoy 29 December 2009 05:36:58AM 0 points [-]

Ah, true.

Comment author: DanielLC 31 December 2009 01:51:35AM -2 points [-]

Always gain mass, once they realized it was negative mass.

The idea that it doesn't always gain mass doesn't falsify phlogiston any more than it falsifies oxygen for the same reason.

Also, people didn't find the change in weight particularly useful, so this wasn't that big a problem.

Again, the vacuum thing isn't much of a problem either. It's not necessarily possible to purify phlogiston.

Comment author: bigjeff5 27 January 2011 05:44:27PM 3 points [-]

I'm not sure I follow, oxidization doesn't predict gaining or losing mass (on any scale like phlogiston would, that is), it predicts an interaction of materials forming a new composite substance. Oxidation doesn't prevent material from being lost or changed in other ways which could cause an overall greater or lesser mass than the original object. What it does predict, however, is that the total mass of all molecules in the equation, once accounted for, will be the same. This is consistent with observation.

If phlogiston has a negative mass, then anything that can burn must gain mass. I don't see any way around it. The theory states that it is a release of negative material, and there is no way to account for it once released.

One thing you would expect to find with phlogiston is an object that was primarily made up of phlogiston, giving it a negative mass. Explosives, for example, clearly have so much phlogiston that it literally rips the object (and anything nearby) apart when released. You would therefore expect all explosives to be relatively light in spite of the original weight of their components.

You could test this with black powder: saltpeter, charcoal, and sulfer each release a certain amount of phlogiston when burned. Combine them and a significantly more phlogiston is clearly released. You would therefore expect more phlogiston to have flowed into the material during the combination of the three objects during the making of gunpowder. However, the weights actually stay quite the same. The observation doesn't bear out the prediction, so the prediction is clearly wrong. If the prediction is wrong, the theory that made it is either wrong outright, or flawed in some way. Since the only prediction phlogiston can make is wrong, then the theory is at the very least flawed in some crippling way, and needs to be completely re-worked.

It's lack of ability to predict expectations is what killed it. You can predict what will happen when you add oxygen to a reaction. You cannot predict what will add phlogiston to a material, thereby allowing it to burn.

A huge example is the sun. It is a giant ball of fire - therefore, a giant ball of phlogiston, or at least a very significant portion of its mass to be made up of phlogiston in order to burn that intensely for that long. So it should have a low mass, possibly even a negative mass. Yet this giant ball of mostly phlogiston is actually the heaviest thing in the solar system by a massive margin.

Phlogiston is incompatible with many, many theories that have been independantly verified. Also, oxygen causing fire is not the theory. The theory is molecules and their chemical interactions, of which oxygen is just one type, and the predictions of oxygen causing most of the exothermic reactions is consistent with all other chemical reactions and is predictable based on rules that are consistent whether a reaction is exothermic or endothermic, among a great many other things. It also predicts which objects will burn and which will not. This same chemical theory leads to atomic theory, which predicts fusion, which has absolutely nothing at all to do with oxygen, yet describes the behavior of the sun very accurately before you even start to measure the sun's output.

The way to test a theory is to predict first, then observe. This is basic science. Phlogiston cannot pass this test, chemical theory can.