Fiction as cultural compression?

The distinction between different types of novels is often very arbitrary—we seem to just randomly assign them into two buckets (“Serious Literature” vs. “Web Novels”) and then compare them. A better approach is to establish a unified framework for comparison.

We can view this through the lens of an autoencoder to understand how humans perceive the world. Our world consists of a massive amount of sensory information ($X$)—what you see, what you hear—and the brain simply cannot process all of it. So, the brain acts as an encoder, compressing this complex information into a conceptual bottleneck ($z$) to help us understand. Just as an autoencoder reconstructs the original image through its bottleneck, we reconstruct life through these conceptual bottlenecks.

These conceptual bottlenecks can be understood as culturally shared beliefs: heroes, villains, leveling up, revenge, childhood sweethearts, and so on. Reading a novel is essentially extracting $z$ from the text ($X$) and then using it to reconstruct $X’$. The nature of this conceptual bottleneck determines the genre of the novel, while the difficulty of that reconstruction process—the inference cost—is, I believe, what truly differentiates them.

Web Novels (Net Lit) have an extremely low inference cost. They rely on $z$ vectors that are incredibly mature and widely accepted (culturally shared)—the underdog counterattack, the hero saving the beauty, justice defeating evil, the sanctity of friendship, hard work paying off. You don’t need to infer what the trope is; the encoding and decoding happen almost instantly. Because these texts rely on such standard $z$, they are very easy to parse into different sections: one chapter, one $z$. Chapter 1 covers the protagonist’s tragic backstory; Chapter 2 covers the protagonist encountering a “cheat” or special opportunity. These stories repeatedly confirm that the reader’s existing cultural values ($z$) are correct, providing a sense of safety and comfort.

Serious Literature changes the dynamic. These novels usually lack a “ready-made” (meaning culturally trendy) conceptual bottleneck. Without relevant literary training, an average reader might find it very difficult to infer a conceptual bottleneck that can reconstruct (explain) the text ($X$) with low loss. You simply cannot infer a valid $z$ to effectively comprehend the full text. You have to re-learn. At the same time, you can’t parse these texts easily; many character motivations are hidden, requiring active inference. Most importantly, these works challenge your cultural values, forcing you to reorganize your belief system (${z}^n$) when your inference fails. The gratification here comes from the reconstruction of meaning.

From this perspective:

• The classic nature of Classical Literature stems from its $z$ being unaffected by time.
• The rebellion of Postmodern Literature comes from the meaninglessness of $X$ itself.

In postmodern works, the author makes it impossible (or very hard) for a modern human to learn a $z$ that can compress $X$. The gratification of postmodernism lies in its meta-nature: you cannot use the standard $X \to z \to X’$ loop to understand the novel. The failure of meaning-making itself is the point, making you doubt your understanding of “understanding novels.”

A side note: creating a truly meaningless $X$ is actually difficult. For example, random input characters still have meaning (“I am inputting random characters”), so postmodernism is hard because dissolving meaning is, in itself, a form of meaning. Therefore, it has to hover between meaning and meaninglessness. Postmodern works provide just enough structure to induce your brain to start looking for $z$ (making you feel like “there’s something here”), but every time you try to establish a concrete $z$ (like irony), the text deliberately introduces destructive structures to invalidate that $z$.

• If it were completely random, the brain would classify it as noise, archive it, and feel pleased.
• If it were completely ordered, the brain would classify it as a story, archive it, and feel pleased.

Postmodernism forces the brain’s autoencoder to keep training, but the loss never goes down—yet it never gets high enough for you to just give up and treat it as pure noise.

Take Waiting for Godot as an example. If the two characters were just speaking pure gibberish, they would be madmen (easily explained/low loss). But they seem to be waiting for someone; there seem to be religious metaphors; they seem to be discussing philosophy. Your brain desperately tries to invoke, say, an “Existentialism Script,” but Godot never comes. That meta-pain is exactly the effect postmodernism seeks to achieve.