Shadow-Loom: Causal Reasoning over Graphical World Model of Narratives
Shadow-Loom is an innovative open-source framework designed to convert narratives into a versioned graphical world model. It utilizes two reasoning engines: one based on Pearl's causation ladder for causal physics and another employing counterfactual calculus over Ancestral Multi-World Networks. Additionally, a narrative physics component evaluates the graph against four reader states—mystery, dramatic irony, suspense, and surprise—drawing from Sternberg's triad of curiosity, suspense, and surprise, alongside computational suspense studies. Large language models are limited to extraction, rendering, and auditing tasks, while identification, intervention, and counterfactual reasoning are executed in typed code on the graph. This system is intended as a research artifact, not a production tool.
Key facts
- Shadow-Loom is an experimental open-source framework.
- It turns a narrative into a versioned graphical world model.
- Two engines act on the model: causal physics and narrative physics.
- Causal physics is grounded in Pearl's ladder of causation.
- Counterfactual calculus uses Ancestral Multi-World Networks.
- Narrative physics scores against mystery, dramatic irony, suspense, and surprise.
- Framework builds on Sternberg's curiosity/suspense/surprise triad.
- LLMs are used only at the boundary: extraction, rendering, audit.
- Reasoning is carried out in typed code over the graph.
- System is a research artefact, not a production tool.
Entities
—