From biodiversity data to knowledge in the age of AI: where do ontologies and standards fit?

What this was about

Ramona Walls, now leading the Data Collaboration Center at C-Path in the biomedical regulatory world, argues that the path from observation to data to evidence to knowledge depends on meaning, trust (provenance, context) and relevance ('context of use'). Using the Information Artifact Ontology's definition of data, examples such as a feather versus a paw print as evidence, and DNA analysis of 'Yeti' specimens, she asks whether AI generates data and knowledge, concluding that it generates data and candidate knowledge that only becomes knowledge when supported by trusted evidence. She stresses the roles of standards bodies, governance ('good governance makes altruism possible') and ontologies (Plant Phenology Ontology, BCO), and offers biomedical models (OMOP CDM and vocabulary, FHIR, European Health Data Space, a C-Path knowledge graph) as potential lessons for biodiversity, concluding that TDWG builds the infrastructure that turns data into evidence.

Why it matters. It reframes TDWG's standards and ontology work as infrastructure for evidence generation in an AI era, and brings an outside perspective from biomedical regulation, where independent, long-lived standards bodies and portable governance determine what regulators will trust.

In the room

• Data are the result of an observation or measurement, almost always a sample, and are intended to represent the truth; this aligns with the Information Artifact Ontology definition of a data entity as a subclass of information content entity.
• The knowledge pyramid places information above data, the opposite of IAO's hierarchy, illustrating that how terms are defined determines how entities are understood within a worldview.
• Transforming data into knowledge requires meaning, trust (provenance, context) and relevance (context of use), which together produce evidence.
• Different kinds of data carry different strengths of evidence (a feather versus a paw print).
• AI generates data (text, images, synthetic data, predictions, classifications, inferred values) but not new empirical observations; it reproduces knowledge and creates candidate knowledge, which becomes knowledge only when supported by trusted evidence.
• Secondary reuse of data is the only way to scale inquiry, so collectors' minimal notes are insufficient without standards and context.
• Regulators such as the FDA trust international, independent, long-lived standards organisations rather than individual academic ontology projects; TDWG (since 1985) plays that role for biodiversity.

Notable moments