Classifying Elements: Groups or Labels?

Groups and labels are optional: an ontology works perfectly without them. They do, however, bring real comfort:

groups make the ontology easier to read and navigate visually
labels let the AI agent and Explore users filter, across the ontology, elements that share a common characteristic

Criterion	Create a group if…	Assign a label if…
Purpose	You want to visually organize nodes to make the ontology easier to read	You want to classify or filter across elements, for the AI agent and in Explore
Scope	Nodes only	Nodes, relations and attributes
Classification	Vertical: each node belongs to a single parent category	Transversal: elements of different types and groups share a characteristic
Multiple membership	A node belongs to one group only	An element can carry several labels. E.g. `BATTERY` carries `safety_critical` and `recyclable`

Example of groups (HR ontology):

Example of a label: the regulatory label can apply at 4 levels, depending on the precision required:

On a whole node — CANDIDATE_PROFILE, because all its information (identity, contact details, application history) falls under labor law and GDPR
On a whole relation — EMPLOYEE — undergoes — BACKGROUND_CHECK, because the background check itself is governed by legal obligations (consent, retention period)
On a node attribute — in PERFORMANCE_REVIEW, only disciplinary_notes is labelled; review_date and manager_id are not
On a relation attribute — in EMPLOYEE — holds_skill — SKILL, only certification_expiry_date is labelled; proficiency_level is not

Elements of different types, and nodes from different groups (PERFORMANCE_REVIEW and CANDIDATE_PROFILE), thus share the same label.