Scenes, Sequels, and Sad Puppies

Why do people like some stories and not others?

While at Worldcon the week before last, I happened upon the poster room: billboards full of posters from the "Academic Track", usually analyzing the theory of story. There was one very interesting poster by Dr. Anthony Francis on "The Cognitive Science of Scenes and Sequels." Although Dr. Francis's talk wasn't quite as interesting as I'd hoped from the poster, it sparked me to thinking.

Meanwhile, in the back of my mind during Worldcon was the "Sad Puppies" battle circa 2014, where the culture war erupted into Worldcon and the Hugo Awards as a fight over whether some sorts of traditionally-conservative stories were being properly appreciated by Worldcon fandom. One of the most cogent thinkpieces I saw back then was by Eric Flint, arguing that it's inevitable that longtime fans of a genre will like different sorts of stories from new fans.

What I realized later, while pondering Dr. Francis's theory, was that these two things feed into each other.

---

Francis's theory looks at stories themselves. Specifically, he looks at one measurement for story quality: "Narrative Transport" without "Narrative Reactance." In other words, does a story "transport" the reader by holding their attention and making a good impact on them, without making them want to throw the book away?

He argues - it's an unproven theory, which he's working to prove using AI analysis of stories - this measure is best served by a succession of "scenes" (with high action) and "sequels" (with low action) tying in with each other. All this does make sense to me and does explain a lot of my own experience reading.

Of course, I'm very doubtful how an AI can test this theory.

Definitely, an AI could definitely judge what are scenes and sequels, at least if fed adequate training data, and graph the relative tie-rods between each scene and sequel. After all, that's what a large language model does: it analyzes large amounts of language. This'd be difficult to train without overfitting on specific authors or styles, but I'm sure it can be done. (One other audience member correctly pointed out that training it on public-domain English-language stories would bias it in itself.)

However, judging how the audience will actually react is different. An LLM can analyze a story; it can't analyze an actual human mind. So, no LLM can reliably predict how a human will react to a scene and sequel. If you're defining goodness of a story by how it achieves Narrative Transport without Narrative Reactance - well, those are phenomena of a human mind, so you need to look at a human mind to see whether they're working.

Perhaps Francis was meaning to cross-check AI results with how people actually react to those stories; if so, that could be a good test.

---

However, even then, people are different. The experiences they bring to the story are different.

During the Sad Puppies fight, one of the most cogent thinkpieces I saw was by Eric Flint. Flint, the author of the 1632 time-travel series – one of the series that in my opinion goes farthest in not deriding the masses – argues that it's inevitable that longtime fans of a genre will like different sorts of stories from new fans. People who've read a lot of stories in a genre, he says, will have seen the commonplace tropes dozens of times before and be tired of them; they'll be looking for something new that plays off their expectations in new ways. Meanwhile, new readers will enjoy the commonplace tropes, because they haven't read them dozens of times, and they'll miss stories that play off longtime readers' expectations.

I think Flint is correct. I've seen it in my own reading.

Eric Flint, in 2007

So, not just any scene and sequel will work. To use Francis's terms, you want something that the readers won't recoil from. You want something which will keep them taking in more of the story at steady speed, without their minds pulling them out of the story. Francis's story-internal analysis of this makes me realize how much it depends on what the reader brings to the story.

---

One book I remember fondly from my childhood is The Great and Terrible Quest. When I reread it as an adult, I thought it was a fair story, but a mere retelling of a very common story of a young boy who gets drawn into a quest for the lost royal heir. Its characters are noble, with a few strikingly-good lines that linger with me, but in many ways they're two-dimensional. To me as an adult, our protagonist's unknown backstory is transparent. But when I was a child, I loved it. The backstory was actually a mystery to me, the characters were amazing, and the story was something I might've seen a few times before but was still fairly new to me.

As Flint says, now I'm a more experienced reader, so the familiar tropes played straight are old to me. I do still enjoy them when explored from new angles in more complicated stories, but The Great and Terrible Quest wasn't that. The scenes and sequels are still there, but they were great to me as a child but not now. I've seen the same thing elsewhere too, in other childhood books and internet fanfiction and elsewhere.

This's why I would advise young readers to read Lord of the Rings before they read much of the modern fantasy genre imitating it. Tolkien was writing in an age where his tropes weren't yet tropes; he played them straight without worrying about reader expectations because readers wouldn't have been expecting anything like it.

If people come from more modern books, they might be tired of these tropes already and not have a chance to experience Tolkien's telling that is richer than many others. Or, even if not, they might easily be suspicious at all the wrong moments and have a weaker experience from having the wrong sort of expectations. Many don't – but many do.

One of Tolkien's many later imitators. A good book in its own right, but it reads better if you're already familiar with the tropes it plays with.

(Perhaps this argues in favor of reading every genre from the beginning forward, moving through history? There'd be a lot of problems with that, but advantages too.)

---

Flint argues, based on this, that the "trufen" of Worldcon (as they name themselves) had read so much of science fiction and fantasy that the old-style tropes enjoyed by the "Sad Puppies" were stale to them, and so the trufen voted for other things in the Hugos rather than the more-popular books the Sad Puppies advocated. This makes a lot of sense to me.

But now, after Worldcon and after pondering Francis's analysis, I think there was another point beyond the Sad Puppies fight too.

There're also other differences in what people like. As Francis himself said, he likes didactic stories but most other people don't. What gives him Narrative Transport doesn't give it to others. But people are more fractally different, even, than that: I'm sure he prefers didactic stories which are didactic about some points more than other points.

Naked politics itself plays a role in this. If you're strongly in favor of one political side, and you strongly believe in its models of reality, you're going to feel more "transport" for a story that conforms to those models. All else equal, an American progressive is going to get more easily into a story where a corporation is the villain than one where it's the hero, and an American conservative is probably going to be the opposite.

That's why the Sad Puppies could point to so many Hugo winners written from one political perspective; that's why I could point this year to how virtually all Hugo finalists were anti-institutionalist. So, sadly, there isn't going to be a politically-neutral method of fully telling story quality.

Fortunately, I'm convinced, there are a lot of ways to tell stories that would be appreciated by both political sides. So, I hope we can ignore this about most individual stories, at least if we read with an open mind like I try my hardest to do. In aggregate, though... the results from groups dominated by different political persuasions can't help looking different.

---

Sometimes, of course, the opposite happens. You like a story so much that you absolutely do want to read it again and again. Sometimes you literally want to reread the same book; other times you want to read more like it; still other times you want to read the same story with different words. The third often leads to internet fanfiction; I've felt it myself. The second led to the explosion of fantasy novels after Tolkien, and is probably at the core of many genre fans today; I've felt it myself too.

A grab-bag of a few books I've reread time and time again

But this isn't in contradiction to the other effect. Rather, it complements it. My taste in fanfiction has absolutely changed since I first started upon it; so has my taste in genre books. Some people take more or less time to get to that point overall... and there'll still be individual works that bring back nostalgia and the same delight time and time again.

---

So, what can we say about scenes and sequels and similar literary qualities that make a story good or bad?

They don't, actually, make it good or bad. Instead, they make it better or worse. Any theory about Narrative Transport and Narrative Reactance analyzed solely in terms of the story itself is going to at best predict a first approximation, and then there will be individual messy variance.

But more likely, it's going to be one element that goes into whether a given person likes a given story, which other elements may often outweigh.

This will pose a problem for testing any such theories. For any one reader, any issues with story structure might be more than countered by other reasons for liking or disliking the story. And if they read two versions of the same story with different structures, they'll likely view the second one differently because they've read it twice. To really test this theory, you'd need to get such a large number of subjects as to drown out individual variation in large numbers – and I don't know enough statistics to estimate how many that would be.

I'm convinced this's the case. Francis has given a major theory of what will make a story better, all else equal. But, all else is very rarely equal.