Skip to main content
An edge goes from a cause to an effect. Draw one by dragging from a variable’s handle to another variable. Click an edge to open its inspector.

​
Edge attributes

  • Arrow type a solid causal arrow (A → B), or a dashed link for a relationship you don’t want to assert as causal.
  • Note a free-text label shown on the edge.
  • Sign (+/−) + (cause up, effect up), − (cause up, effect down), ? (unknown), ~ (non-monotonic).
  • How strong? small / medium / large.
  • How do you know? experiment / data / theory / guess.
  • Confidence a slider from hypothesis, through observation, to strong.
  • When does it apply? an optional condition, e.g. “only when ad ROI ≥ 1.5×”.
  • Source where the claim comes from, e.g. “WHO 2020 report”.

​
Review

Edges start as pending review. Mark one reviewed when you’ve checked it: reviewed edges are preserved even when LLM auto-restructuring runs over the graph.

​
The DAG constraint

The graph must stay acyclic. Drawing an edge that would create a cycle (A → B → A, directly or transitively) is rejected with an explanation: causal graphs don’t allow arrows that loop back. If two variables genuinely influence each other, that’s usually a sign you need a third variable both depend on, or a time-lagged split of one of them.