7. Parsimony

Parsimony means frugality or simplicity. Within philosophy and science, parsimony has taken on a special meaning, distilled in a heuristic (or guideline) known as Occam’s Razor. The gist of Occam’s (or Ockham’s) Razor is that one should not posit more entities to explain a phenomenon than are necessary. Overly complicated explanations tend to be inaccurate descriptions of past phenomena and poor predictors of future phenomena. One’s explanations should therefore be “frugal” or parsimonious. If you really want to understand why and in what circumstances a parsimony heuristic is useful, you may want to wade into a book-length discussion, such as philosopher Elliott Sober’s 2015 book Ockham’s Razors: A User’s Manual. My analysis will be simpler. I won’t get into the theories of probabilistic inference that inform a sophisticated answer to these questions. Instead, I’ll offer several concrete examples that give you a sense of the relevant contexts and uses of parsimony heuristics. First, however, I caution that parsimony is not the kind of rule that admits no exceptions. There are an infinite number of wrong explanations for every phenomenon, many highly parsimonious. Thus, there is no guarantee that a highly parsimonious explanation is correct, nor that an unparsimonious explanation is incorrect. Moreover, sometimes people inappropriately appeal to parsimony to justify explanations that are overly simplistic. In predictive modelling, this problem of oversimplification is often called “underfitting,” and it is just as problematic as unparsimonious, or “overfit” models. Another kind of oversimplification can happen with reductive explanations, if they fail to acknowledge emergent phenomena. In future videos, I will discuss oversimplification, including underfitting and problematic forms of reductionism (though not all forms of reductionism are problematic). Today, I’ll be covering the opposite of oversimplification, showing how certain forms of simplicity (or parsimony) can provide a mental heuristic that is useful in diverse contexts. I describe three of these contexts below. Examples of where parsimony heuristics are helpful (and exceptions): 1. Critiquing Theoretical Constructs—e.g., Conspiracy Theories. At first, a conspiracy theory may seem perfectly parsimonious. The explanation is that the official view is wrong, and we are being deceived by those in power. Seems parsimonious enough, right? However, conspiracy theories quickly become unparsimonious when they attempt to explain away real-world data. Consider the QAnon conspiracy theory. Each time a “Q-drop” turns out to be wrong, QAnon followers invent ad hoc explanations purporting to show that Q wasn’t really wrong. As a result, the QAnon conspiracy involves a great number of distinct hypotheses. Taken individually, some of these hypotheses may seem somewhat plausible. However, if you consider all of them at once the likelihood shrinks down to virtually nothing. Compared to the alternative hypothesis—that Q is a fraud—the Q conspiracy is ultimately so unparsimonious as to be totally implausible. Similarly, consider the “flat earth” conspiracy. Each time “flat earthers” are presented with a strong argument against their claim that the earth is flat, they must invent an ad hoc explanation that preserves their hypothesis. As a result, believing in the flat earth conspiracy forces one to accept a large number of implausible claims. For instance, it is claimed that the government (NASA, the military, etc.) is involved in a conspiracy designed to trick the public. That already feels like a stretch—but the government does lie to us, right? So…maybe? Well, to preserve the flat earth hypothesis, one must also accept that private industry (airlines, GPS tech, SpaceX, Blue Origin, etc.) must be part of the conspiracy. And, of course, pilots in general. And the entire internet must be monitored and censored so that photos and videos that show the flat earth are taken down, and instead search engines are flooded with doctored images and videos of a globe. And academia—physicists and astronomers are either fools or they’re part of the conspiracy. Ditto for the mainstream media. All these hypotheses and more must be accepted to preserve the flat earth hypothesis. Additionally, new hypotheses must be fabricated to explain, for instance, lunar eclipses. Or the fact that the sun does not appear twice as big at noon as it does in the morning and evening. Or the fact that ships viewed through a high-powered telescope eventually disappear over the horizon. Or that different constellations are observed on different hemispheres—and they also change with the seasons. Speaking of which—the seasons themselves require a wholly new explanation. And so on. Or one could simply accept the more parsimonious thesis that the earth is a large three-dimensional globe that orbits the sun. (Essay continued in the comments)