Chesterton's fence and modern life
How do we navigate new developments as our environments grow farther from those of our ancestors?
You may be familiar with Chesterton’s fence already, but here’s a brief exploration in case you aren’t. The English writer G.K. Chesterton criticized reformers who wished to tear down existing regulations without understanding their purpose by likening the regulations to a fence. If someone uncovered the reason the fence had been put up and still wanted to tear it down, that person’s understanding indicated that it was much safer for him to tear down the fence (though Chesterton didn’t claim that understanding was sufficient to ensure the tearing down was justified).
Chesterton’s fence has often been invoked to defend traditional customs with no apparent rational purpose. A confident young person might want to discard such customs, but in doing so, she should be wary that she didn’t hurt herself or her society. One particularly striking example of an apparently nonsensical tradition that actually served an important purpose is the intricate process that indigenous tribes in the Colombian Amazon used to prepare manioc. This plant is a tuber that contains cyanide in high enough levels that it’d eventually kill you if you kept eating it without preparing it properly, but low enough levels that it’d take decades to do so, thus making the link between the preparation and the poisoning murky. As Scott Alexander said in the linked post, a reasonable person not thinking about Chesterton’s fence wouldn’t see any reason to spend so much time preparing manioc and would thus eventually poison herself.
Chesterton’s fence is a useful reminder of how useful traditional wisdom can be in the context in which it was developed. But given how quickly our society is changing, how can we be so sure that wisdom applies today? Indeed, we sometimes see an implicit nod to Chesterton’s fence when people advocate for ending societal prohibitions on certain types of behavior. One example is when people acknowledge that homophobia might have served a useful purpose when humanity was much less populous and benefitted from population growth, but now no harm is done by people being in same-sex relationships and not reproducing, since the Earth is so populous.
Unfortunately, tracing the sources of our world’s actual Chesterton’s fences isn’t as simple as wandering around the town with the fence and asking residents why the fence was put up. Even the homophobia example is largely speculative; we don’t know for sure that cultures adopted homophobia to encourage people to ignore their same-sex attractions and have kids for the sake of humanity. Prohibitions that are more unique to particular cultures are even murkier. Sure, maybe the Jewish prohibition on pork developed to protect the ancient Jews from trichinosis, and modern Jews with access to fancy stoves can safely fry bacon. But maybe there’s some substance in pork that slightly raises one’s risk of certain negative health conditions, and we’re still struggling to do good enough nutrition science to find it.
We’re less and less able to rely on the wisdom of our elders, and the accelerating pace of technological change makes it imperative that we find new ways to guide ourselves. Perhaps we could get some assistance by constructing a Taylor series around older wisdom. (Note: the rest of this paragraph is a brief explanation of the Taylor series, so feel free to skip if you’re familiar with it.) A Taylor series is a way to approximate a function near a certain region. If you know the value of sin(0)
and the rate of change of sin(x)
at 0, you can estimate values like sin(0.0001)
reasonably well. However, the Taylor series centered at 0 is a poor approximation if you want to compute something like sin(1000)
. Note that the graph below (source and license) yields another interesting insight. (The sine function, which is being approximated, is in black.) As the degree of the approximation increases, meaning that more information about the function’s trends is taken into account, the approximation remains accurate for longer. The degree 1 polynomial, plotted in red, is a quickly a poor fit once you venture outside of the region near x=0
. But the fit stays good longer for the degree 3 polynomial (orange), longer still for the degree 5 one (yellow), and longest for the degree 13 one (pink), which is the highest degree polynomial in the graph.
So how exactly can we use these insights to guide our choices? Well, if we think of a function whose inputs are variables indicating the current state of a society (including its technological developments) and whose outputs are the optimal way to live for a given societal state, we can see some conclusions. (One limitation is that ancestral wisdom wasn’t necessarily the true optimal way to live even under the conditions in which it was developed, but that’s a topic for another post.) We want to use a sort of Taylor series approximation based on our function’s value (ancestral wisdom) at a previous time (under the societal conditions in which it was developed), but we’re moving farther away from the region where we constructed the approximation (societal conditions are changing more and more). So unless we develop some new wisdom to re-approximate the “optimal living” function at a point closer to where we are, we only have one option. We need to understand more about that function’s longer-term trends around that point, so we can construct a higher-order Taylor series.
One way we can understand those higher-order trends more effectively is by studying history. If we know more about what had happened recently when our elders developed their wisdom, what was about to happen, how quickly those changes were happening, how fast the rate of change was changing, etc., we could estimate those higher-order terms fairly well. Just knowing the facts about our ancestors’ then-current state isn’t enough; we need to understand how they got there and where they went next.
One thing I find a little intimidating about studying history is that I don’t know where to start. If I want to learn more about a particular time period, I feel that I need to go back further than the beginning of the period itself to understand the trends that led there. But it’s often unclear to me how far back to go, and then I also think that if I want to understand the trends leading into the time period, I need to learn about the trends that led to those trends. When I took history classes in school, my teachers solved this problem by being thoughtful about the curriculum and using their broad knowledge to summarize the salient trends from before the time period we were studying. And even now that I’m done with school, I can solve this problem by reading books by historians who’ve come up with similar solutions. But the solution I’m suggesting for the higher-order Taylor series problem is going to be more complicated than packaging up temporal trends into a book. We’ll need to understand how the “optimal living” function changes as some variables that have been constant for much of human history change.
Note (11/19/21): A reader pointed out that some of what I wrote about using a Taylor series to approximate societal choices was vague. After some contemplation, I agree with this criticism. I’m still trying to figure out how to include technical concepts in my writing in a clear way without confusing people who are unfamiliar with those concepts, but I’m already thinking of a follow-up to this post that clarifies what I meant. I’ll try to make that post as accessible as possible, but it’ll probably at least assume the reader is familiar with single-variable calculus. My apologies to those readers who haven’t taken calculus, but I think this post suffered from avoiding being explicit about functions and formulas. In the future, I’ll probably write a mix of posts that don’t require much quantitative knowledge and posts that explicitly state what knowledge they require. In the latter case, I’ll take advantage of the reader’s assumed knowledge to state formulas and definitions clearly.