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

allenpaltrow 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.

Comment author: allenpaltrow 03 June 2011 05:43:50PM *  0 points [-]

I think that this is really a discussion of explanatory power, of which scientific causation is one example. All theories attempt to explain a set of examples. Scientific theories attempt to explain causation in natural phenomena, thus their "explanatory power" is proportional to their predictive power. A unified theory of forces at the planetary and subatomic levels would explain more examples than any do now, thus it would have great explanatory power.

Yet causation isn't the only type of explanatory relationship. Causation implies time and events, whereas these are only one type of explanation. For example, the Pythagorean theorem explains why physical right triangles in reality have the lengths that they do. It doesn't "cause" them to have the properties they do. It would be foolish to say that any property of physical triangles "explains" or "proves" the Pythagorean theorem, because mathematical truths exist independent of practicalities. Plato's dialogue The Euthyphro beautifully explains why even if the set of things which are x and the set of things which are y are equivalent (in that case, the set of pious actions and the set of god loved actions,) they are not the same quality if one (god loved) explains the other (piety) and not vice versa. Similarly, the total number of hydrogen atoms in a glass of water is always even, but it is the quality of evenness (any number which is a multiple of two must be even) that explains this, not any quality of hydrogen. The one "explains" (but does not "cause") the other.

Thus, I think some parts of this post would be better understood as being stated as thus: any theory which provides no additional explanatory power should be ignored.

So, looking at the case of Phlogiston, the OP is not saying it is "wrong," but that it lacks the explanatory power that justifies it as a useful theory. If I take the Neils Bohr model of the atom, and say that there are extra invisible subatomic particles, and that these particles are "god," you would be hard pressed to prove me wrong. But this theory does not predict any new phenomena, nor is it falsifiable, nor, most importantly, does it have an explanatory relationship with any other known truth about atoms: none of them explain this theory, and it explains none of them. It exists completely independent from any other aspect of atomic theory, thus it lacks any explanatory power as a theory.

Yet there are theories which have great explanatory power but not empirical predictive power. Lets say I'm a simplistic deontologist who says that killing is wrong because human life is good. Along comes a utilitarian who says, I have a theory which explains, in all the cases where you're right, why you are right, and in those cases where you aren't, why you aren't, according to your own first principle. In terms of my very simplistic ethical theory, the utilitarian would absolutely be "less wrong" than me, for he has provided a theory which better explains the hard cases my theory failed to (justified killings, kill 1 save 2 etc.)

In the case of the post-utopian author, I think that we again are getting wrapped up in "prediction" when we should concern ourselves with explanation.

What is a plumber? Is it a man who comes to your house, sits on your couch, eats your food, watches your TV, and flirts with your wife? Even if this is true of all plumbers, it is not the definition of plumber. Definitions should be proscriptive, such that they give you the means to determine what counts as an x, and what a good x is. If a plumber fixes pipes, anyone who fixes pipes is a plumber, a good plumber fixes them well, and no one who doesn't fix pipes is a plumber.

Thus, hold literary labels to the same standard. Don't ask, "is this label true"? Because as we saw earlier with the god particle example, many theories cannot be proven false but still have greater or lesser explanatory power (see economics, ethical theories etc). The better standard is explanatory power. Is there a definition of the quality "post-utopian" such that any book with quality x is post-utopian, x explains why it counts as post utopian, and the more x it is, the more it is post-utopian it is? Saying post-utopian is a,b,c,d,e,f,g,h, but failing to provide a single explanation of the aforementioned form is like calling the plumber a man who eats your food and flirts with your wife: it is a descriptive definition, not a proscriptive definition. It may be true of the every plumbers, but it is not the thing that makes plumbers count as plumbers.

I think the OP meant to say that literary labels like post-utopianism fail to meet this standard. Sure, you can come up with descriptive statements of the terms which may be true (post-utopian books do not portray utopian societies as possible) but this is not a definition because it is not this quality that a. makes post-utopian books count as utopian, b. without which a book cannot be post-utopian, and c. designates a clear set of books which either are, or are not, post-utopian. Textual analysis perhaps can be more wrong and "less wrong," but literary theories are just not the sorts of truth-bearing statements that mathematical, scientific, or philosophical theories are.

Compare "post-utopian" to "even". Even numbers are a set of specific numbers, but there is a single quality they have (being multiples of 2) which explains why they are in the set. Without that quality, they would, "by definition", not be even. This is the standard we should be looking for in definitions and theories. Not just that they are "true" (plumbers do steal your food, watch your tv, and flirt with your wife) but that they have the sort of explanatory power we've isolated.

Thus, I think the larger point of the post stands. There are better theories and worse theories, and we should prefer the better ones.

Comment author: Alicorn 03 June 2011 06:00:36PM 1 point [-]

deontologist who says that killing is wrong because human life is good.

Aaaaaaaaugh.

Comment author: allenpaltrow 03 June 2011 06:44:18PM *  0 points [-]

I'm not trying to define the terms, just posit a very very simple theory of the form killing is wrong because human life is good. Such a theory would be inferior on its own premises than a very very simple utilitarianism, regardless of whether either or the premise itself is true. As such I oversimplified utilitarianism just as much, but it doesn't matter for the scope of the example.

Edit: in fact, for the purposes of the example it is better if the "deontologist" is wrong about deontology, because it better illustrates how one theory can have greater explanatory power than another only on the grounds of the former's justification without reference to external verifiability. "human life is good" is a poor first principle, but if it is true, the utilitarian's principle applies it better than the "deontologist's" did.

Comment author: Alicorn 03 June 2011 06:53:14PM 0 points [-]

Someone who believes that killing is wrong because human life is good is not a deontologist. See here.

Comment author: allenpaltrow 03 June 2011 07:32:25PM 0 points [-]

Here the deontologist is arguing for the principe 'killing is wrong regardless of the consequences' (deontic) but uses a poor justification for which consequentialism is a more reasonable conclusion. So the 'deontologist' is wrong even though his principle cannot be externally verified. I was just (unclearly I see) using this strawman to illustrate how theories could be better and worse at explaining what they attempt to explain without being the sorts of things which can be proven. I will attempt to be clearer in future.