The Triple A Principle: actionability, applicability and auditability for actionable biodiversity knowledge

What this was about

Lars Vogt argues that FAIR data and machine-actionability are insufficient for actionable knowledge in interventions such as restoring a degraded mangrove site. He proposes three nested criteria - actionability (procedural knowledge of what to do), applicability (explicit context conditions under which it works) and auditability (evidence of where it succeeded or failed, feeding back into conditions) - and a representation using the Semantic Units Framework: action units, applicable action units and audited applicable action units, plus conditional action units with question and directive units that turn a knowledge graph into a context-sensitive decision support system. The framework is conceptual, with a preprint on arXiv.

Why it matters. It pushes knowledge-graph design beyond describing what is the case toward supporting decisions and interventions such as ecosystem restoration.

In the room

• Many "FAIR" datasets have FAIR containers (metadata) but not FAIR content.
• Machine-actionability usually means data transformation; human actionability includes epistemic and intervention operations.
• Actionability: procedural knowledge (directives, steps toward a goal) is largely absent from FAIR infrastructures.
• Applicability: actionability does not equal applicability due to context dependencies; explicit applicability conditions are needed.
• Auditability: applicability reports feed back to refine conditions (e.g. salinity outside documented success range).
• The three As are nested and proposed as necessary and sufficient for the representational side of actionable knowledge.
• Semantic units make statements first-class typed, addressable units; action units link input, output, plan and objective units.

Notable moments