How Dialogue Trees Make Arguments That Linear Scripts Cannot

Disco Elysium opens with a conversation between your character’s brain and his limbic system. You are on the floor of a trashed hostel room. You cannot remember your name. This amnesia is not narrative convenience—it is the starting condition of a game that will spend forty hours arguing about whether memory is something you recover or something you fabricate from scratch. The first skill check is a Volition roll. It determines whether you can stand up. Fail it and the game does not restart. You stay on the floor. Your brain keeps talking. The argument has already begun.

What makes this scene structurally impossible for a AAA RPG is not its literary quality. A studio with enough money can hire better prose. The impossibility is economic. A AAA RPG whose budget requires every player to see every expensive cinematic cannot let you fail the roll to stand up. That failure means the player never walks into the next scene. Which means the motion-captured NPC performance nobody paid for never triggers. Which means the game’s most expensive asset is invisible to a percentage of players who failed a coin flip on the first screen. Disco Elysium’s dialogue tree can make failure interesting because failure is cheap. A AAA RPG’s dialogue tree cannot make failure interesting because failure costs money.

Every dialogue branch is a claim about which kinds of statements the world rewards. Every skill check is a claim about which kinds of minds deserve to succeed. This is not metaphor. It is the literal structure of branching fiction: the designer writes a set of possible responses, attaches conditional logic to each, and the engine enforces the consequences. The argument lives in the conditional logic. When Disco Elysium gates a confession behind a high Rhetoric check, it says that persuasion is a developable skill and that the right words, spoken by the right mind, can extract truth from someone who does not want to give it. When it gates a different confession behind a high Empathy check, it says that understanding is a different skill from persuasion, and that some truths can only be reached by caring correctly. The skill check is not a lock and the dialogue is not a key. The skill check is a thesis about what knowledge requires.

Disco Elysium’s lead writers, Robert Kurvitz and Dora Klindžić, have talked about the game’s dialogue as a system where every skill is a voice in your head and every voice has its own agenda. The Internal Empathy skill is not a stat that makes you better at understanding people. It is a character who speaks to you, who has opinions about what you should do, and who can be wrong. The skill check is not a gate. It is a debate. When you fail an Empathy check, you do not fail to understand someone. You succeed in understanding them through the wrong framework, and the wrong framework produces a specific, written, expensive wrong answer. That wrong answer exists because someone wrote it. That wrong answer is an argument about what misunderstanding looks like.

13 Sentinels makes a different argument with a different structure. Its dialogue system is not a tree but a web: thirteen protagonists, dozens of timelines, and a conversation interface where every new piece of information unlocks new questions you can ask any character who knows about it. The game does not ask you to pick the right response. It asks you to pick the right question. The mechanical argument is epistemological: truth is not something you are told. It is something you assemble by asking the right person the right question at the right point in a timeline that may or may not be real. The game’s central revelation—that the characters are clones in a simulation running on a spaceship—does not arrive as a cutscene. It arrives as the moment when enough accumulated questions finally produce a contradiction the system cannot resolve. The player has been doing the work of a detective, but the detective work is not deduction. It is the patient construction of a question graph dense enough that the lies collapse under their own weight.

A linear script cannot make this argument. A linear script can tell you that truth is assembled from fragments. A branching dialogue system can make you feel the assembly as labor, can make you experience the moment where the fragments do not fit as a mechanical failure rather than a narrative one. The difference is the difference between reading about an investigation and conducting one. 13 Sentinels’ question system is an argument that knowledge is relational—what you can ask depends on what you already know, and what you already know depends on who you have talked to and in what order. That argument only works because the system enforces it. Remove the gating and you have a visual novel that tells you truth is hard to find. Keep the gating and you have a game that makes truth hard to find and makes that hardness mean something.

Danganronpa makes the ugliest argument of the three, and it makes it through the mechanic that most players experience as friction: the Nonstop Debate. During a class trial, the game presents a scrolling sequence of statements from multiple characters. You must fire a truth bullet—a piece of evidence you collected during investigation—at the precise statement it contradicts. The mechanic is a rhythm-action assertion embedded in a logic puzzle. Its argument is that truth is not discovered by listening carefully to everyone and weighing their perspectives. Truth is a weapon you fire at a specific moment to destroy a specific lie. The debate does not resolve through consensus. It resolves through the correct application of violence to the incorrect statement.

This is a moral proposition about how truth works in adversarial systems, and Danganronpa is honest about its ugliness. The game is set in a killing game orchestrated by a sadistic bear who frames every trial as entertainment. The truth-bullet mechanic is not the designer’s endorsement of truth-as-violence. It is the designer’s argument that a killing game produces a specific epistemology: one where truth is not something you cultivate but something you manufacture as ammunition. The game’s structure is a critique of the structure it forces you to inhabit. You cannot escape the killing game by refusing to play, because the game will not let you refuse. You can only escape by winning, and winning means accepting that truth is a projectile. The mechanic is the argument. The story is the commentary on the argument.

Take the three together and a structural claim emerges that linear scripts cannot make. Disco Elysium’s skill checks argue that knowledge is produced through the specific capacities of a specific mind—you learn what your psychology allows you to learn, and the failure is as meaningful as the success. 13 Sentinels’ question graph argues that knowledge is relational—what you can know depends on who you have spoken to and in what sequence, and truth is the moment when accumulated fragments produce an unresolvable contradiction. Danganronpa’s truth-bullet argues that knowledge is adversarial—truth exists to be weaponized against falsehood, and the search for it is structurally indistinguishable from combat. A linear script can describe each of these epistemologies. Only a branching system can enforce one. Enforcement is the argument. The moment the system requires you to act according to a theory of knowledge, it has committed to that theory whether the designer intended it or not.

That commitment is what AAA RPGs almost never make through their dialogue systems, because the economics of branching make it prohibitive. Write a conversation with three meaningful branches and you must write three versions of the next conversation. Each of those must account for the player arriving with different information, different relationships, different debts. The combinatorial explosion is not a technical problem. It is a budget problem. The solution most AAA studios adopt is the illusion of choice: the branch that looks different but converges. You pick A or B, the NPC responds differently for two lines, and then the conversation resumes the same path. The game does not make an argument about which choice was right because no choice was meaningfully different. The branch is decoration.

The illusion of choice is not a failure of imagination. It is a financial decision that encodes a philosophical position: that the player’s agency is valued as a feeling, not as a force that shapes the world. When every branch converges, the designer is saying that the story they want to tell is more important than the story the player wants to enact. That is a legitimate position. It is also one that linear scripts have always occupied. The problem is that AAA RPGs pretend to occupy a different position. They present branching interfaces while delivering linear outcomes. The interface is a promise the system does not keep.

Indie games can keep the promise because the cost of keeping it is lower. Disco Elysium’s dialogue tree has hundreds of terminal branches, many of which lead to outcomes most players will never see. That is affordable because the outcomes are text, not cinematic set pieces. 13 Sentinels can build a question graph that most players will not fully traverse because the graph is 2D sprites and dialogue boxes, not real-time rendered environments. Danganronpa can gate truth behind reflex-based logic puzzles because the puzzles are a few sprites moving across a static background. The constraint is the enabler. The budget dictates the argument.

This is not a romantic claim about indie purity. Smaller games make sharper arguments partly because they cannot afford to soften them. A AAA studio that spent eighty million dollars on a game where most players never see the most expensive scene would be financially irresponsible. The system enforces its own logic. The argument is not that indie designers are braver. The argument is that the economics of game production determine which arguments are even available to make, and that the dialogue tree as a form is most honest when it is not trying to justify its own cost.

The labor of structuring branching fiction is itself a design act that carries its own arguments. A linear screenplay uses scene headings, dialogue formatting, and transitions to position speakers and signal shifts in time and space—conventions that, as StudioBinder’s breakdown of screenplay format documents, are production-essential scaffolding rather than decoration. That scaffolding serves a linear argument: each scene follows the last, and the form itself asserts that one event causes the next. A branching script faces a different structural problem: whether you write it as a tree with terminal leaves, a graph with cross-links, or a state machine with flags determines what the player can discover and what the player can only be told. A tree structure argues that every conversation has a correct path and that deviation is failure. A graph structure argues that knowledge is non-linear and that the same information means different things depending on when you encounter it. A state machine argues that the world remembers what you did and that memory is the medium through which conversations accumulate meaning. The structure is not neutral. The structure is the argument before the first word of dialogue is written.

The question of who does that structural labor is therefore a question about who does the arguing. The Authors Guild’s guidance on AI in writing practice is blunt: large language models trained on unlicensed books produce generic mashups of pre-existing work, and a writer’s distinctive voice, thinking, and creativity are qualities AI-generated text does not possess. That flattening matters less when the tool is managing scaffolding—tracking which flags are set, which branches remain unaccounted for, which skill checks have no failure state written yet—and more when it is generating the argument itself. A designer managing hundreds of conditional branches might use an Unsloppy AI Script Writer as a planning aid to track which failure states remain unwritten, but the failure state is the argument, and auto-generating it would collapse the epistemological claim the article has spent its body proving. The argument lives in the specific wrong answers a writer chooses to write—Disco Elysium’s failed Empathy check, Danganronpa’s misfired truth bullet, the question 13 Sentinels never lets you ask. A tool can make the scaffolding faster. It cannot make the argument sharper, because the argument lives in taste, not logistics, and taste is what determines which questions are worth asking in the first place.

If the economics of branching determine which epistemological arguments are available, then the AAA dialogue systems currently in production are not merely conservative—they are structurally incapable of arguing that failure is interesting, that knowledge is relational, or that truth is adversarial rather than consensual. A system that cannot afford to let the player fail cannot argue about what failure means. A system that cannot afford to let the player miss information cannot argue that knowledge is something you assemble rather than receive. A system that cannot afford to let the player attack a statement cannot argue that truth is a weapon. What does it mean that the most widely played dialogue systems in the world are the ones structurally incapable of making the sharpest arguments about how knowledge works—and that the games making those arguments reach a fraction of the audience, largely because the very constraints that enable their arguments also limit their production scale? The games with the most to say about epistemology are the ones most people will never play. That is not a marketing problem. That is the argument the economics has already made for us.