Looking at Society Quantitatively

Distributions, decomposition, and fundamental parts

Current social science, including mainstream HBD thought, uses quantification in a way which is ancillary to words. For example, a very common technique in the HBD sphere is the data review blogpost, which involves gluing together data sources with a logic that is ultimately verbal.

The fundamental downside of this approach is that you can often infer things which are not logically implied, and that therefore turn out to be untrue. That’s because verbal thinking is heuristic. Sometimes the heuristic has a positive truth EV, and sometimes it doesn’t. In either case a heuristic is never perfect, so it guarantees nothing. Quantitative reasoning, in contrast, is completely logical. What is inferred from it is known with certainty. This is a necessity for wading into counter-intuitive areas.

The upside to verbal thinking is that it is easy and efficient. The downside to quantitative thinking is that it is difficult, and when applied to the wrong use cases, it provides nothing new over the verbal approach. But I think there are some topics where the quantitative approach is necessary and very beneficial.

Those topics include memetics, social change and behavioral evolution, and social power. To study these topics, I developed some quantitative methods, and I would like to detail the ideas behind explicitly in this article, so that hopefully they can be useful to somebody else.

Individuals as samples from population behavioral distribution

The approach I used to study social change and behavioral evolution started with probability and statistics. It’s too hard to predict individual behavior, but what about mass behavior? It turns out this should be theoretically easy. If individual deviations from the population mean cancel out, then you don’t have to worry about them; you can simply study how the population mean evolves over time. It turns out this is quite quantitatively tractable with evolutionary probability & statistics.

Decomposition of phenotypic pressure/velocity

Now that were are studying population means, we want to explain how they change. We do this with decomposition. It’s analogous to decomposing the forces acting on an accelerating particle in physics.

From phenotypic longitudinal surveys you can measure that the phenotypic mean for leftism changes 0.20 SDs over some decades, for example. Now you break this down into types of forces acting on it. My idea was to do genetics, memetics, and other environment, as something of an analogy to ACE decomposition in twin studies.

So the sum of the forces dG + dM + dE should add up to dP = 0.20 in this example. We immediately see that we only have 2 degrees of freedom; if we can measure two of these terms were get the third. Additionally we can use a simpler equation with two terms, genes and the rest of environment. The extended environment is the sum of “cultural” or “memetic” change and all other types of possible environmental change.

To measure these, we would like them to be as well defined as possible. It turns out dG is already well defined and it’s easy to extend the infinitesimal theory of dG to dM as well. This leads the the next main ideas.

Stripping down phenotypic evolution forces to atomic components

dG at its core is driven by changes in allele frequency, which is mathematically described by the theory of population genetics. dG is a long sum of constituent atomic components that have weights. The theory of regression solves the weights. The weights are the average effect of a change in gene frequency on the phenotype under study.

I realized this model is quite elegant and could also apply to the memetics compoment dM. I read a couple of books on modeling cultural evolution, and they did not really get this; they were filled with intractable differential equations modeling the spread of individual memes. The central idea of these books was that you should model macro memes, like the meme Christianity, and assume that when someone adopts “Christianity”, their entire behavior changes in a saltatory fashion. Thus, you need to track the spread of the macro meme with differential equations, like in viral epidemiology. It is unclear how this should interact with the infinitesimal model of genetics when applying it to the study of polygenic traits.

These models were produced largely in the 1980s and never went anywhere. It seemed that they existed to be pointed to by anti-hereditarians as a kind of road block, “look, here we have cultural evolution, have you dealt with this in your work? go read it, it’s important.” But it was like neoclassical economics in that the models were completely without estimable parameters. They could not be used in real life.

So I decided a better model would be an infinitesimal model of memetics, with macro memes coming as a special case of this, like how Mendelian traits are actually a rare special case of the polygenic model. People don’t adopt macro memes whole sale, they are constantly being blown around by small memes that have microscopic effects on behaviors, and that might add up over time to produce larger effects. Religion is a special case of this model where closed off communities end up with different memetic averages than others due to heterogeneity in the spatial distribution of memes. When that heterogeneity breaks down, you have to use the more general infinitesimal model to get anywhere, because everyone is exposed to dozens of different religions that all vary in a clinal fashion from one another on top of being flopped around by a bunch of other memetic forces that all cancel out to some average.

In theory can you give people large surveys of ideas, asking if they have ever been exposed to this or that bit of information, then you can try to correlate that with their phenotypes. This will produce estimates of the effects of those ideas on them. I was also able to work out some parameters for interactions with genetics. For example, people with different genotypes may tend to produce different types of memes, meaning genetic evolution causes memetic evolution, which can then amplify the genetic evolution. Memetic clustering can be caused by people with different genotypes accepting different types of memes more. Thus the difference between religious groups probably has at its base a genetic origin, although memetic spatial heterogeneity can serve to amplify behavioral differences between such groups.

Applying these ideas to social power

To analyze social power in society, I thought again to use decomposition. That’s the secret formula. What I ended up with was breaking down the structure of society into the elite class and the warrior class. It would seem all social power comes from people with elite positions in the economy, government, or ideological arena, plus people who enforce dictates from such people which is the warrior class.

I then thought about the moral or political substance of laws and determined that, in a perfect democracy, by definition they should follow the median of the population, which under a bell curve is also the mean. And due the the infinitesimal structure of the underlying forces creating that mean, it is in fact a bell curve. So social power really only matters with respect to moral policy if the policy deviates from that mean. Why would it do that? Well, because the elite or the warrior class deviate. So this lends itself to a parameter: elite deviance and warrior deviance. It would seem that warriors today are pretty average, and elites lean a little bit left, but not that far left. It depends on the specific institution. The government does not really lean left very much.

What produces this deviance? Probably population stratification. Likely genetic in origin, resulting from different evolutionary forces acting at different levels of society.

And what is an elite, anyway? In theory we can break it down quantitatively using the techniques I’ve already outlined. If you could measure policy changes in society against people becoming a certain age or dying, you could potentially construct a measure of how influential any given person is. Then you can correlate these scores with their traits. I hypothesize that IQ correlates positively with such scores, as well as some other traits. I expect that it mainly comes down to SES which captures both monetary wealth and some ideas of softer power, though I expect money to matter somewhat more than the latter.

Conclusion

Here I’ve outlined the main techniques I’ve used to quantitatively model society. I view society statistically and I think there’s great potential in users statistical parameters as the measurable observables in models. Mainly these are regression weights, and the regressions come from performing quantitative decompositions and reasoning about the basic parts of whatever aspect of society is being measured. Usually things can be linked all the way back to alleles. Otherwise, one can reason about infinitesimal memes or enviremes.

While I’ve measured some aspects of what I’ve outlined here, such as evolutionary pressures, I haven’t been able to measure memetics or social power parameters due to lack of funding. Maybe that will change in the future.