Dragna's Blog: Stories are about people, not plots

I. STORIES ARE ABOUT PEOPLE, NOT PLOTS

There's a quote from Game of Thrones said by Lord Stannis Baratheon, one of multiple candidates vying to win the throne of Westeros by conquest. After a notable defeat, he journeys north to the frozen wastes beyond the Wall, where the Night's Watch is currently trying to resist a zombie invasion. His decision is fueled by a single realization: "I was trying to win the throne to save the kingdom, when I should have been trying to save the kingdom to win the throne."

Stannis does not, ultimately, wind up winning the throne (though he does help save the kingdom). But I've gone through a similar sort of revelation in the past seventy-two hours: "I was trying to build the characters to serve the plot, when I should have been trying to build out the plot to serve the characters."

Let me explain what I mean.

II. HUMANS ARE SOCIAL CREATURES

Seven months ago, I recruited a group of friends (who are new D&D players) for a brand-new homebrew campaign. We played on and off for a few months, with folks dropping in and out in an adventure-of-the-week, West Marches-style format. Then, two months ago, they told me that, although they wanted to keep playing their characters in the same world, they wanted to transition to a more regular, more serious campaign with backstories, themes, an overarching plot—the works.

And so I got to work. I started fleshing out lore, a campaign-level dramatic question, and ideas for seeds and sprouts that I could sow to set up future narrative arcs. I built a world map and developed settlements, waterways, politics, histories, and natural features. I divided the campaign into six acts, then divided the first act into four arcs, each of which would take place in a different settlement. I got player backstories and started figuring out how to work them into the lore.

I designed what I felt was a kickass first "reboot" session, complete with fun skill-heavy games, familiar resonant NPCs (both beloved and beloathed), and a big, climactic inciting incident to kick things into gear.

My players loved it. But to me, it felt flat, sloppy, and lame.

The scenes had dramatic questions, technically—but they felt hollow, forced, and half-baked. The NPCs felt like paper cutouts—exposition monkeys half the time, and generally nonresponsive to the players' actions. If not for my own experience running games (which allowed me to riff and spin off the players' reactions to keep energy high and the players engaged), it would've been an absolute flop of a session.

So I went back to the drawing board. I tried to figure out where I'd gone wrong, and soon decided that I just hadn't put enough work on. I hadn't done enough "prep," as the saying goes—and I'd gotten what I should've expected from my lackluster efforts. I'd been trying to get a Reloaded-quality product without the benefit of Reloaded amounts of time, a Reloaded-level existing world, or Reloaded-quality existing experience with the product. I wasn't willing to compromise on the quality of the campaign I wanted to run, but I couldn't spare the time to dump hours upon hours of time into "prep." 

So I took the only alternative option: I decided to cheat.

I started by feeding summaries of campaign design theory into ChatGPT, and then asked it to flesh out narrative ideas for the arcs and acts to come. I was a good designer and editor, I reasoned, but a slow creator at best—it’s my personal belief that I have a void in my brain where my muse should be. So, I figured, why not create my own co-DM to do all of the worldbuilding and inspiration-hunting for me?

It went decently well at first—if there's one thing ChatGPT is good at, it's coming up with ideas—but I soon hit a wall. The first, and most obvious, issue was that ChatGPT (even GPT-4) isn't always the best at reasoning. It'll do what you tell it, but the further you push it into areas that require lots of strategic, conscious design instead of creative, flexible prose, the more it'll fall into stubborn patterns and struggle to follow your train of thought.

That wasn't the biggest problem, though. As I used it to churn through idea after idea, I felt a sinking, empty feeling. All of the ideas it gave me were functional—like, they could be used as the basis for a D&D narrative. But none of them resonated. I remembered the feeling I'd had when drafting Stella's return in Reloaded's The Lost Soul, or the Martikov family's dramatic strife in The Stolen Gem—that lump in my throat that told me yes, this is something pure and good and special that speaks to the heart—and that feeling was totally absent here.

The narrative ideas that ChatGPT was giving me weren't anything special. They weren't even particularly memorable. They were just grist for the adventuring mill.

To be clear—I don't begrudge any DM who's content to run "grist." The word isn't meant as an insult—many joyful and memorable campaigns have come more from the players' spontaneous decisions and ideas than any complex narrative the DM can prepare. But it didn't fit for me.

Disgruntled and anxious, I sat down for an evening of television—specifically, an episode of the new live-action One Piece Netflix series. (I've been a One Piece fan since middle school, and the show had, so far, incredibly exceeded my originally rock-bottom expectations.) Watching the series start from the beginning of the protagonists' adventure had rekindled my love for the characters and their world, and the stories that defined them.

For those who don't know, the first arc(s) of One Piece focus more on the protagonist, Monkey D. Luffy, meeting and recruiting the first people who will form the backbone of his pirate crew. These initial arcs have to do a lot of legwork to set these new shipmates up, introducing and developing their backstories, giving them meaningful early-game character arcs, and giving them a reason to travel with Luffy and devote themselves to him as a captain.

Many of these are tearjerking arcs. They have antagonists, to be sure, and big, climactic fights—but the stakes are so dearly intimate and personal. I'd felt that lump-in-my-throat effect on multiple occasions so far, and tonight's episode was no exception.

It only when the credits were rolling, however, that I realized something: None of these introductory arcs were, ultimately, about their narrative dramatic questions—narrative dramatic questions that, for the most part, were permutations of "Can Luffy defeat the evil pirates before they do bad things and hurt the local villagers?"

Instead, these arcs were, ultimately, about the characters that Luffy was recruiting—their dreams, their pain, their struggles, their flaws, their setbacks, and their breaking points.

To put it another way, it was clear that the author hadn't started with an idea for a narrative arc and then asked, "How can I make some characters that will be involved with and care about this plot?" Instead, it seemed clear (to me, anyway) that the author had started with a character—vivid, dynamic, and complex—that he wanted to join Luffy's crew and then created an arc designed to deeply explore raw and intimate aspects of their character.

Defeating the villain of each arc wasn't important because it meant that the good guys won. Defeating the villains was important because it meant that the protagonists and/or their friends obtained emotional catharsis and a new revelation about their relationships and/or place in the world.

If you look across One Piece's overall runtime, the author does the same thing consistently, again and again. Now, once the series exits its first act and the crew is mostly already recruited, this trend does shift somewhat and it becomes more evident that the author started with a narrative first and then fleshed out a broader cast to support it. But every arc, without fail, the author writes the story such that the audience's primary investment comes from the emotional stakes of the characters, not the dramatic stakes of the narrative.

To put it another way, the most important and memorable moment of each arc isn't the moment when the protagonists beat the villain. It's the moment just before that—when a character achieves self-actualization amidst the darkest depths of despair, and makes the decision to continue forward nevertheless.

Everything after that? Is just falling action.

III. THE SECRET TO BEING LAZY

So here's what I did next:

I took my NPC profile template. I gave it to ChatGPT, along with the bare minimum amount of information I knew was set in stone for the arcs to come. I told it the bare minimum about the NPC I wanted to flesh out. I told it what kind of resonance I wanted from that character (if any). Altogether, aside from the template, I gave it little more than a short paragraph of material to work with.

And then I . . . set it loose.

And as I read what it wrote—on the first try—I felt that lump in my throat coming back.

It's often said that humans are a social animals. For the most part (although there are exceptions), we care more about stories about us—our triumphs, our downfalls, our hopes, and our despairs—than about stories that focus the world around us. And as I read what ChatGPT had written—which, might I note, wasn't anything particularly special, innovative, or unique!—I felt the little tug of emotion swelling in my little primate forebrain.

It didn't get everything right on the first try, of course. These AI models work best when you can have a conversation with them—tossing ideas and edits back and forth in a natural dialogue—and, especially on subsequent tries, my narrative plans were too ambitious for it to hit the nail on the head immediately.

But the emotional, character-driven narratives it gave me were more than enough fuel to keep me going—more than enough of a flame to light my way through the dark. Within fifteen minutes, I had not one, but four NPCs (plus related arcs) that I felt confident about fitting into my campaign—and each one had reliably (almost) made me tear up. (If you’re interested, you can read my conversations with ChatGPT here, here, and here.)

Despite ChatGPT's limited memory window (only about 3,000 words or so), it was easy to spin my campaign's lore into the plans "we" were weaving, working more complicated narrative seeds into raw character arcs. But my success soon sent me thinking—how had I been able to create resonant arcs in Reloaded without using this method? After all, I'd always started with the narrative first, and fleshed out the characters second—including when making the revised guide.

And then I realized: I hadn't started with the narrative—because I hadn't been working from a blank slate. After running the Curse of Strahd campaign three consecutive times and writing hundreds of thousands of words about it, I knew its characters better than some people know their friends. Every time I developed a new narrative idea, I was unconsciously sculpting it around what I already knew and believed about the campaign's NPCs—from Van Richten’s desire for redemption to Izek's monstrous obsessions, and even Strahd's hubris and cunning.

I often say to DMs, "Don't make content for the sake of content." Whenever you're designing something—be it an encounter, a magic item, or a twist reveal—make sure that it serves the broader purpose of the narrative and gameplay experience.

Now, I think I've got another aphorism to add to the pile: "Don't make plot for the sake of plot." Our players think they're coming to the table for a high-stakes story with dragons to slay and treasure to loot. But what they don't know is that, deep down, many of them crave more: a story that will tug at their heartstrings, featuring friends and foes that they'll remember for a lifetime.

Start with the characters—NPC and PC alike. Start with the journeys they'll go on: the trials they'll face, the dreams they'll pursue, and the lows to which they'll fall. Figure out how they'll make your players roar with delight, rage in hatred, and sob in bittersweet melancholy.

And then—only then—put pen to paper and figure out the stories that will make them shine.