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

RogerS comments on Reductionism - Less Wrong

40 Post author: Eliezer_Yudkowsky 16 March 2008 06:26AM

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

Comments (150)

Sort By: Old

You are viewing a single comment's thread.

Comment author: RogerS 23 April 2013 04:24:57PM *  0 points [-]

"having different descriptions at different levels" is itself something you say that belongs in the realm of Talking About Maps, not the realm of Talking About Territory

Why do we distinguish “map” and “territory”? Because they correspond to “beliefs” and “reality”, and we have learnt elsewhere in the Sequences that

my beliefs determine my experimental predictions, but only reality gets to determine my experimental results.

Let’s apply that test. It isn’t only predictions that apply at different levels, so do the results. We can have right or wrong models at quark level, atom level, crystal level, and engineering component level. At each level, the fact that one model is right and another wrong is a fact about reality: it is Talking about Territory. When we say a 747 wing is really there, we mean that (for example) visualising it as a saucepan will result in expectations that the results will not fulfil in the way that they will when visualising it as a wing. Indeed, we can have many different models of the wing, all equally correct - since they all result in predictions that conform to the same observations. The choice of correct model is what is in our head. The fact that it has to be (equivalent to) a model of a wing to be correct is in the Territory. In short, when Talking about Territory we can describe things at as many levels (of aggregation) as yield descriptions that can be tested against observation.

at different levels

What exactly is meant by “levels” here? The Naval Gunner is arguing about levels of approximation. The discussion of Boeing 747 wings is an argument about levels of aggregation. They are not the same thing. Treating the forces on an aircraft wing at the aggregate level is leaving out internal details that per se do not affect the result. There will certainly be approximations involved in practice, of course, but they don’t stem from the actual process of aggregation, which is essentially a matter of combining all the relevant force equations algebraically, eliminating internal forces, before solving them; rather than combining the calculated forces numerically.

...the way physics really works, as far as we can tell, is that there is only the most basic level—the elementary particle fields and fundamental forces

The way that reality works, as far as we can tell, is that there are basic ingredients, with their properties, which in any given system at any given instant exist in a particular configuration. Now reality is not just the ingredients but also the configuration - a wrong model of the configuration will give wrong predictions just as a wrong model of the ingredients will. The possible configurations include known stable structures. These structures are likewise real because any model of a configuration which cannot be transformed into a model which includes the identified structure in question is in conflict with reality. Physics is I understand it comprises (a) laws that are common to different configurations of the ingredients, and (b) laws that are common to different configurations of the known stable structures. Physicalism implies the belief that laws (b) are always consistent with laws (a) when both are sufficiently accurate.

...The laws of physics do not contain distinct additional causal entities that correspond to lift or airplane wings

True but the key word here is “additional”. Newton’s laws were undoubtedly laws of physics, and in my school physics lessons were expressed in terms of forces on bodies, rather than on their constituent particles. The laws for forces on constituent particles were then derived from Newton’s laws by a thought experiment in which a body is divided up. In higher education today the reverse process is the norm, but reality is indifferent to which equivalent formulation we use: both give identical predictions.[Original wording edited]

General Relativity contains the additional causal entity known as space-time curvature, which is an aggregate effect of all the massive particles in the universe given their configuration so is not a natural fit in the Procrustean bed of reductionism. [Postscript] Interestingly, I've read that Newton was never happy with his idea of gravitation as a force of attraction between two things because it implied a property shared between the two things concerned and therefore being intrinsic to neither - but failed to find a better formulation.

The critical words are really and see

Indeed, but when you see a wing it is not just in the mind, it is also evidence of how reality is configured. It is the result of the experiment you perform by looking.

.. the laws of physics themselves, use different descriptions at different levels—as yonder artillery gunner thought

What the gunner really thought is pure speculation of course, but this assumption by EY raises an important point about meta-models.

In thought experiments the outcome is determined by the applicable universal laws – that’s meta-model (A). In any real-world case you need a model of the application as well as models of universal laws. That’s meta-model (B). An actual artillery shell will be affected by things like air resistance, so the greater accuracy of Einstein’s laws in textbook cases is no guarantee of it giving more accurate results in this case. EY obviously knew this, but his meta-model excluded it from consideration here. Treating the actual application as a case governed only by Newton’s or Einstein’s laws is itself a case of “Mind Projection Fallacy” – projecting meta-model (A) onto a real-world application. So it’s not a case of the gunner mistaking a model for reality, but of mistaking the criteria for choosing between one imperfect model and another. I imagine gunners are generally practical men, and in the field of the applied sciences it is very common for competing theories to have their own fields of application where they are more accurate than the alternatives – so although he was clearly misinformed, at least his meta-model was the right one.

[Postscript] An arguable version of reductionism is the belief that laws about the ingredients of reality are in some sense "more fundamental" than laws about stable structures of the ingredients. This cannot be an empirical truth, since both laws give the same predictions where they overlap so cannot be empirically distinguished. Neither is any logical contradiction implied by its negation. It can only be a metaphysical truth, whatever that is. Doesn't it come down to believing Einstein's essentialist concept of science against Bohr's instrumentalist version? That science doesn't just describe, but also tells? So pick Bohr as an opponent if you must, not some anonymous gunner.