Graphonauts: Introducing a Hybrid Human-AI Pipeline for Text-to-Game Transformation via Knowledge Graphs

Abstract

This study represents the first systematic attempt to transform ethnographic texts into playable game experiences using knowledge graphs (KGs) as an intermediary framework. While large language models (LLMs) promise automated KG construction from text, dense ethnographic writing resists full automation: its interpretive richness requires human expertise to preserve nuance, ensure transparency and maintain scholarly integrity. We therefore introduce a pioneering two-stage hybrid workflow that integrates expert interpretation with computational assistance. In the first stage, a domain expert collaborates with an LLM to construct a knowledge graph, guiding the model to extract salient concepts and relationships from ethnographic source material. In the second stage, a custom tool procedurally renders this KG as a playable 2D platformer where concepts become traversable platforms and relations appear as interactive mechanics such as ladders, bridges, and navigable gaps. Using the anthropological classic Argonauts of the Western Pacific as a case study, we analyze through autoethnographic analysis how players experience what we call experiential hermeneutics, a fusion of epistemic, kinaesthetic and interpretive pleasures, when moving through theoretical structures rather than merely reading them. We argue that this hybrid workflow offers a pragmatic and reflexive path for integrating LLMs into playable scholarship while foregrounding the interpretive and affective dimensions of game-based knowledge representation.