Identity through roles and change
Separate what something is, the capacity in which it participates and the changes it can survive. Make the identity commitment visible before choosing a role or phase encoding.
Technical modelling guide Modelling judgement Chapter 03
Method basisUpdated
Maya sells one dwelling while buying another. A report about her sale is corrected. A proposal would subdivide the dwelling. All three situations involve change, but they do not ask the same identity question. A model that treats them as one “status” will lose the differences that matter.
One person, two participations
In the fictional Harbour Court case, Maya participates as seller in transaction A and buyer in transaction B. Identify Maya independently of either participation. Then identify the relationship in which each capacity holds. There is no contradiction in the simultaneous roles, and ending one participation need not change Maya’s identity.
A timeless isSeller flag cannot say which transaction it concerns. Two dates stored directly on Maya do not reveal whether they bound selling, buying or something else. The defect is not that booleans and dates are weak datatypes: the model has omitted the relationship to which the facts belong.
A qualified participation can identify the bearer, the capacity and the transaction together, with its effective period and evidence where those are needed. Giving that participation a record identifier does not give Maya a second identity. It distinguishes the participation from other participations by the same bearer.
Now change the scenario. Maya is eligible to act in a capacity but never participates. The model must be able to represent that without asserting an actual transaction role. ODR-0042 requires independently expressible eligible-role and played-role relationships; a list of possible capacities cannot stand in for actual participation.
- A building may remain Building B while its occupancy or condition changes, provided the adopted identity criterion supports that continuity.
- A later record can revise an earlier description without changing the building or the historical event it describes.
- Subdivision, amalgamation or demolition require an explicit identity review. This illustration does not settle whether a new entity has begun.
- The diagram shows logical sequence, not a measured time scale.
Kind, subkind, role and phase do different work
A kind supplies the relevant identity criterion. A subkind specialises it rigidly while retaining that criterion. A role is an anti-rigid classification dependent on a relational context: the bearer can cease playing it while remaining the same individual. A phase is an anti-rigid classification associated with intrinsic change in an individual taken to persist.
These distinctions are not determined by a category’s English ending, its number of fields or how long it has been recorded. A sparse role can have few attributes and important relationships. A label that sounds temporary may conceal a different identity. An age-related classification and a seller role may both cease to apply, but one concerns an intrinsic condition and the other a relationship.
OPDA’s selected encoding uses rdfs:subClassOf for genuine local subkinds and governed roleOf/phaseOf annotations for class-level role and phase relationships. Those annotations describe a type’s relationship to a bearer type. They are not standard W3C predicates and do not themselves record a particular person’s dated participation.
This is a local modelling policy. The OntoClean rigidity constraint prohibits an anti-rigid superclass from subsuming a rigid subclass; it does not universally prohibit a role from being a subclass of a kind. Keep the analytical rule, the chosen encoding and the instance-level fact separate.
- Person is the illustrative identity-supplying kind. The actual criterion still requires domain justification.
- Buyer is an illustrative role grounded in a relevant external relation; someone may cease being a buyer and remain the same person.
- An age-defined adult classification is an illustrative phase grounded in an intrinsic condition under an explicit classification criterion.
- The local class-level annotations Buyer roleOf Person and Adult phaseOf Person identify bearer types; they are not rdfs:subClassOf edges or instance-level participation assertions. The individual retains the bearer’s identity under its justified criterion, rather than acquiring a new identity from role or phase membership.
Put a fact where its meaning begins
Ask of each proposed property: would it still describe the bearer outside this role or phase? An enduring party identifier describes the party. The period of Maya’s participation in transaction A describes that participation. A review outcome concerning one report does not become a property of every person mentioned by that report.
| Proposed fact | Its subject | Why the distinction matters |
|---|---|---|
| Maya’s name | The person | It does not cease to make sense when the sale ends. |
| Seller participation in transaction A | The scoped participation | It can coexist with a buyer participation elsewhere. |
| Participation’s effective period | That participation | The dates must not qualify every role the person plays. |
| Report was corrected | The report version relationship | A correction is not another inspection. |
| Transaction entered a reviewed state | A transition concerning the transaction | The report and transaction have different state spaces. |
| Flat 1 will be subdivided | A proposed change requiring identity analysis | A state record cannot decide which dwelling persists. |
The introducing-class rule does not mean every role needs a bespoke property for every fact. Reuse relationships where their semantics genuinely fit. What it rejects is hoisting role-specific meaning onto a general bearer simply to shorten a diagram or reuse a database table.
A state value is not automatically a phase class
A transaction lifecycle might use a closed set of states agreed for that context. A report has a different lifecycle. Both may contain a label such as “under review”, without sharing the same subject, transition rules or effective period. A global Status scheme can erase those differences while making the vocabulary look tidy.
A controlled state value can be sufficient when the requirement is to classify and track a persistent entity. A phase type is a stronger analytical choice about how that entity is classified under intrinsic change. Do not create phase classes for every workflow label, or assume that every operational status meets the phase definition.
Where the adopted transition pattern applies, a transition record identifies the subject, previous state, new state and effective instant. The two values come from the same governed state space and differ from one another. A dedicated allowed-successor relationship describes one-step adjacency. skos:broader remains a conceptual hierarchy, not a synonym for “happens next”.
The state record says what changed, not why a bearer is the same individual before and after. That identity commitment must already be defensible. The time chapter adds the distinction between when the state applies and when its record was created.
The hard case: what survives a subdivision?
Suppose Flat 1 is to become two dwellings. One proposal keeps Flat 1’s identifier for the larger successor. Another retires it and introduces two new identities. A third follows a documented administrative continuity rule. Each can produce a well-formed graph. Syntax cannot choose among them.
Ask each proposal to state its criterion and consequences. Would it give the same answer if the larger successor changed? What happens to earlier inspections of the original extent? Which descriptions concern a former whole and which concern a successor? Can an address remain while the relevant individual changes? Which responsible context owns that definition?
Do the same for demolition, replacement, merger and boundary change. These are challenges to a proposed criterion, not five pre-agreed answers. The historical schema-derived model and a current candidate may embody different assumptions; neither becomes the domain’s adopted resolution merely by existing in a corpus.
A useful unresolved result names the competing interpretations, their evidence and the decision owner. An unqualified “subdivision is a phase” hides the question by presuming persistence. An unqualified “always create new resources” hides it in the opposite direction.
Make validation follow the intended subject
A class-level role annotation does not inherit a bearer’s SHACL constraints automatically. Give each applicable class an explicit target and its own shape under ODR-0042, including sparse roles. Compose justified repeated checks through the selected shape profile; do not invent sh:extends or rely on an unexamined chain of shapes.
Test three cases together: an actual seller participation with its transaction, an eligible party with no participation, and the same bearer participating in two different transactions. The first should receive the participation checks, the second should not acquire an invented participation, and the third should preserve distinct qualifiers. These tests reveal modelling errors that a single populated “seller” example cannot expose.
Separately test the declared metaproperties and hierarchy through four-axis analysis. Passing the participation shape does not prove rigidity, identity, dependence or unity. The tests assess different commitments and need separate evidence.
Decision basis
ODR-0041 governs kinds, roles, phases and class-level annotations. ODR-0042 governs introducing-level properties, sparse roles and explicit shapes. ODR-0053 governs state records; ODR-0059 governs cross-context identity comparison; ODR-0061 supplies the analytical checks.
These method requirements do not certify the fictional example or current candidate implementation. Harbour Court’s subdivision and demolition criteria remain deliberately unresolved.
Comments
Loading comments…
Sign in to post a comment