Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CALYPSO: LLMs as Dungeon Masters' Assistants

Andrew Zhu, Lara J. Martin, Andrew Head, Chris Callison-Burch

TL;DR

The paper tackles the cognitive burden on Dungeon Masters in Dungeons & Dragons and presents CALYPSO, a trio of LLM-powered interfaces (Encounter Understanding, Focused Brainstorming, and Open-Domain Chat Baseline) that distill context and brainstorm within a campaign setting. It reports a large-scale, in-the-wild study with 71 participants, showing that LLMs can function as synchronous co-DMs by delivering high-fidelity content suitable for players and low-fidelity ideas that preserve DM creative agency. The results indicate that explicit prompts to apply thematic commonsense enable richer, narratively coherent material, while preserving human oversight and agency. The work argues for a paradigm of AI-augmented tools in tabletop gaming and outlines future directions toward multimodal inputs and tighter tool integration to further reduce friction in live play.

Abstract

The role of a Dungeon Master, or DM, in the game Dungeons & Dragons is to perform multiple tasks simultaneously. The DM must digest information about the game setting and monsters, synthesize scenes to present to other players, and respond to the players' interactions with the scene. Doing all of these tasks while maintaining consistency within the narrative and story world is no small feat of human cognition, making the task tiring and unapproachable to new players. Large language models (LLMs) like GPT-3 and ChatGPT have shown remarkable abilities to generate coherent natural language text. In this paper, we conduct a formative evaluation with DMs to establish the use cases of LLMs in D&D and tabletop gaming generally. We introduce CALYPSO, a system of LLM-powered interfaces that support DMs with information and inspiration specific to their own scenario. CALYPSO distills game context into bite-sized prose and helps brainstorm ideas without distracting the DM from the game. When given access to CALYPSO, DMs reported that it generated high-fidelity text suitable for direct presentation to players, and low-fidelity ideas that the DM could develop further while maintaining their creative agency. We see CALYPSO as exemplifying a paradigm of AI-augmented tools that provide synchronous creative assistance within established game worlds, and tabletop gaming more broadly.

CALYPSO: LLMs as Dungeon Masters' Assistants

TL;DR

The paper tackles the cognitive burden on Dungeon Masters in Dungeons & Dragons and presents CALYPSO, a trio of LLM-powered interfaces (Encounter Understanding, Focused Brainstorming, and Open-Domain Chat Baseline) that distill context and brainstorm within a campaign setting. It reports a large-scale, in-the-wild study with 71 participants, showing that LLMs can function as synchronous co-DMs by delivering high-fidelity content suitable for players and low-fidelity ideas that preserve DM creative agency. The results indicate that explicit prompts to apply thematic commonsense enable richer, narratively coherent material, while preserving human oversight and agency. The work argues for a paradigm of AI-augmented tools in tabletop gaming and outlines future directions toward multimodal inputs and tighter tool integration to further reduce friction in live play.

Abstract

The role of a Dungeon Master, or DM, in the game Dungeons & Dragons is to perform multiple tasks simultaneously. The DM must digest information about the game setting and monsters, synthesize scenes to present to other players, and respond to the players' interactions with the scene. Doing all of these tasks while maintaining consistency within the narrative and story world is no small feat of human cognition, making the task tiring and unapproachable to new players. Large language models (LLMs) like GPT-3 and ChatGPT have shown remarkable abilities to generate coherent natural language text. In this paper, we conduct a formative evaluation with DMs to establish the use cases of LLMs in D&D and tabletop gaming generally. We introduce CALYPSO, a system of LLM-powered interfaces that support DMs with information and inspiration specific to their own scenario. CALYPSO distills game context into bite-sized prose and helps brainstorm ideas without distracting the DM from the game. When given access to CALYPSO, DMs reported that it generated high-fidelity text suitable for direct presentation to players, and low-fidelity ideas that the DM could develop further while maintaining their creative agency. We see CALYPSO as exemplifying a paradigm of AI-augmented tools that provide synchronous creative assistance within established game worlds, and tabletop gaming more broadly.
Paper Structure (38 sections, 4 figures, 1 table)

This paper contains 38 sections, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Background and Related Work
  3. Dungeons & Dragons in the Time of COVID
  4. Large Language Models and D&D
  5. Design Motivation
  6. Inspiration.
  7. Strategic Copilot.
  8. Thematic Commonsense.
  9. Simple Language.
  10. Implementation
  11. Encounter Understanding
  12. Summarization.
  13. Abstractive Understanding.
  14. Focused Brainstorming
  15. Open-Domain Chat Baseline
  16. ...and 23 more sections

Figures (4)

  • Figure 1: After rolling a random encounter (red), DMs can use LLMs with Calypso to help generate an encounter scene and digest information about monsters. Calypso can present monster information concisely (green) and brainstorm conversationally (purple) to help build a compelling narrative to present to players (purple).
  • Figure 2: Using the Encounter Understanding interface to generate a distilled summary of an encounter.
  • Figure 3: DMs found the Abstractive Understanding method of distilling monster information more consistently helpful than the Summarization method.
  • Figure 4: Using the Focused Brainstorming interface to ask specific questions about an encounter. Calypso suggests reasons why the players might encounter the monsters and how they might act.