“Underfitting and Overfitting” Versus “Right and Wrong”

Using terms from computer science to help us think through historical development.

Aristotle was a genius, and yet I think Aristotelian logic today, mainly in the form of “A = A,” must be negated/sublated into the “A = B” logic of Hegel, so in what sense do I mean it when I say, “Aristotle was a genius?” Was he really? Yes, absolutely, because relative to the time he wrote he offered the world an incredibly useful model of thought that “practically worked” for thousands of years. There really was no (immediate or practical) need for Hegel, and arguably had Hegel been introduced to the world before Aristotle, no one would have had an idea what Hegel was talking about and even “rightly” discounted his work as useless. Something similar can be said about Newtown before Modern Physics and many other examples: the point is that times change, and yet it also seems too much to say that “Newtown was wrong.” Yes, Newtown and Aristotle were “incomplete,” but I think “wrong” is a little much. However, that designation can feel like all we have to use given the restraints of our current language.

To be clear, I am not saying there is no such thing as “right and wrong,” that it’s not acceptable to say that “Aristotle is wrong” in x or y way, or even to say “Aristotle was wrong” regarding an “overfit”-situation—my point is that there is risk with the language and that we should always use terms like “right and wrong” (in this context at least) alongside an understanding of “fittingness.” Graham Harman uses the language of “underfitting and overfitting” to describe the ontology of things, but here I want to use the language to help us understand thinkers and ideas in the history and development of thought (though I might admittedly not be using the terms exactly the same way as they are used in Computer Science, so forgive me if I err in that manner). Furthermore, I want to stress that we must go through Aristotle to understand Hegel, as Newtown is necessary for Einstein, so discussing one as better than the other feels inappropriate. To help make the point for why, imagine that you went back to 1200 and started telling someone about YouTube. Obviously the person wouldn’t understand what you were talking about, but I want to stress how deeply the person wouldn’t understand the topic. YouTube requires a person to know about the internet, videos, cameras, lighting, electricity, science, modern economics, infrastructure—the list of prior knowledge necessary to grasp something like “YouTube” is so vast that I hardly know where to begin. And yet for us it would probably feel as if the idea of “YouTube” was really simple—why couldn’t this person in 1200 “get it?” Well, because over the years all our “background knowledge” becomes so habitual and integrated into our daily lives that we don’t even realize it is there (a Heideggerian point), and thus it’s almost impossible for us to really get why this person in 1200 struggles so much to understand the topic of “YouTube.” However, as we move forward in history from 1200 to 1300, 1400, 1500, 1600…the difficult effort might slowly and incrementally prove easier (until the amazing jump from 1900 to 2000).

Understanding a given thing often requires incredible background knowledge that is mostly “invisible” to us most of our lives, and if we simply lack that background, gaining it might be impossible, until we undergo the experiences necessary so that we might gain it. Knowledge of a thing is rarely if ever just a matter of “foreground” or “directness”: we require context, definitions, and long chains of information that we gathered over many years and decades which make it seem as if all we need is foreground, but really this is a mistake. There is a background at work, hidden behind our foreground.

Here, we can start to understand why thinkers like Aristotle and Newtown are utterly necessary for us to today understand and appreciate contemporary thought across fields. Unfortunately, I fear the language of “right and wrong” and thinking of Aristotle as “wrong,” for example, contributes to us failing to appreciate or understand the role of “background” and historical development in making contemporary thought possible. Furthermore, there are times when it would simply be easier and “practically correct” enough to use Aristotle instead of Hegel, say when all we need is a simple identity to make a linear conclusion about a linear situation, topic, etc. But if we think “Aristotle is wrong,” that might not strike us as a possible angle we can take without betraying or threatening truth.

All that said, we can start to suggest why it is better in my mind to say Aristotle “fit” for thousands of years up to arguably recently, perhaps 2000 (even if Hegel wrote Science of Logic in 1810s, for Hegel may have been early), though now still ascribing to Aristotle (in some areas) is “overfitting” Aristotle, which is to say we attempt to use Aristotle to do too much. Logic which assumes linear and identity (“A = A”) does not help us understand emergent Global Systems, even if Aristotle might “fit well enough” regarding a given individual to his or her individual life. Just to make a simple example, on very large scales and very small scales, Aristotelian logic will prove “incomplete,” as it will if we try to use it to define an entity according to its entire unfolding or development through time (or “becoming,” as Hegel describes). In such examples, Aristotle isn’t wrong but “overfits” and/or “proves incomplete,” though do note that we can find parts of Aristotle in Hegel, which is further reason to avoid saying that Aristotle “was wrong.” The problem though is that we cannot so readily find Hegel in Aristotle, hence why we must “historically develop.”

To “overfit” a thinker is to try to make his or her work apply to phenomena that it cannot explain without reducing that phenomena, simplifying it beyond recognition, or the like. Now, problematically, if I use Aristotle to explain “emergence” or “becoming,” I easily don’t see what I “leave out” about emergence because I use this framework, and as a result the consequences of my mistake are easily and probably invisible to me. This suggests a reason why the stakes here are so high: if we fail to adjust our thinking, the consequences of that failure can be concealed from us, which means we might not have reason to think we failed at all. And so we might fail further and see even less…

Alright, but when do we “underfit” a thinker? A good question, and I should note that “overfitting” tends to be the mistake we make, because all we have to bring into the present is what we have from the past, and if what has worked thus far has worked thus far, there’s reason to believe we should keep using the same model, even if it’s time to try something. That said, for me, “underfitting” would be to use M-Theory to explain the curvature of a ball thrown into the air, or to use Hegel to explain my sister’s linear and immediate identity. Sure, M-Theory and Hegel can be used for these purposes, but it’s also to apply both to something of a complexity level that can be explained better and more simply with Aristotle and Newtown. I would be using something complex to describe something simple without adding value through the extra work; sure, if making something simple complex created more value, that would be fine, but in this circumstance it doesn’t. In this way, we “underfit” Hegel when we use him to explain something Aristotle explains well-enough for our purposes (note the phrase “our purposes,” for this suggests “fitted-ness” is relative to intentions), which basically means that “underfitting” is a matter of using something complex to explain something simple, when a simpler model would be more “fitting” (please note that by “simple” I do not mean “less valuable” or “less important,” because where simplicity applies we are given a great gift). In general, “overfitting” is when we use a model to explain something which exceeds the scope of the model, while “underfitting” is using a model to explain something that a scope of a simpler scope would explain just as well (and “Philosophical Developmentalism,” as we’ll explain in The Absolute Choice, should be understood with these terms). We seek “fitted-ness,” but of course that begs the question of how. Indeed, how…

In closing, we require thinkers of the past who are now “incomplete” to us in order to have the historic “background” we require today to experience and discover new ideas in our “foreground,” which is to say that if we understand Hegel today it is partially thanks to Aristotle—in what sense is it fair to say, “Aristotle was wrong,” then? But all this means we are vulnerable to “overfitting” and “underfitting”—a necessary price so that we might think anew. The language of “right and wrong” can contribute to us not recognizing or optimally handling this “historic developmentalism” which we require to think, whereas “fitting, overfitting, and underfitting” might help us keep in mind that we need our history to think our present (furthermore, the language feels easier to associate with Hegel’s language like sublation, negation, etc.). If we can interpret our present, we have hope to gain the vision with which the future can shine.

.

.

.

For more, please visit O.G. Rose.com. Also, please subscribe to our YouTube channel and follow us on Instagram, Anchor, and Facebook.