Use Case 2 · Explainability

Generating Plan Explanations

Challenge

Planners produce correct plans but little human-readable rationale for what happens and why.

Ontology-driven approach

Run SPARQL over the plan knowledge graph, then fill results into natural-language explanation templates.

User survey: ontology-generated explanations rated highly across three explanation types (avg out of 5).

SPARQL · high-level plan summary

SELECT ?step ?action ?object WHERE {
  ?plan :hasStep ?step .
  ?step :hasAction ?action ;
        :actsOn   ?object ;
        :stepNumber ?n .
} ORDER BY ?n

Explanation →

"First unstack B from A, then place A on B and B on C to build the goal tower."