Skip to main content

argumentation.dev

Formal argumentation for scene AI — auditable reasoning, interpretable beats

Inspectable scene traces

Each beat records the arguments asserted, the attacks that bound, and the residual that produced the acceptance — something you can read, replay, and reason about.

60+ Walton schemes out of the box

Argument from expert opinion, analogy, cause-to-effect, slippery slope — all as composable scheme instances with premises, conclusions, and critical questions.

β — scene intensity as a single parameter

Tune how strictly attacks bind. Low β — every counter bites (courtroom energy). High β — counters slide off (boardroom cordiality). One parameter, a wide range of scene registers.

Trait-inverted encounter bridge

StateActionScorer and StateAcceptanceEval plug into any consumer's scene engine through the encounter crate's trait-inverted interface.

Based on published research

Primitives from Dung (1995), Walton-Reed-Macagno (2008), Cayrol & Lagasquie-Schiex (2005), Dunne et al. (2011), Modgil-Prakken ASPIC+ (2014). Every type traces back to a paper we link.

Rust-first, WASM-ready (soon)

Workspace of small, composable crates. Zero unsafe. Deterministic by default. A WASM build is on the roadmap for browser-native demos.