Originally published September 21, 2023. Updated to clarify the developing status of STAMP and its relationship with JSON-LD, CTDL, talent taxonomies, and the Human Intelligence Layer.

Organizations have more information about people, learning, and work than ever before.

Résumés contain employment histories. Learning systems record course completion. Credential platforms issue badges and certificates. HR systems maintain roles and performance information. Project platforms capture activities and evidence. Skills systems extract labels from job descriptions and employee profiles.

Yet combining these records rarely produces a coherent understanding of human capability.

The problem is not simply that data is stored in different places.

It is that different systems use different structures, definitions, assumptions, and levels of context.

One system may describe “leadership” as a skill. Another may treat it as a competency made up of several behaviors. A school may define a learning outcome related to collaborative decision-making. An employer may use “cross-functional influence.” A professional organization may issue a credential in team leadership.

These concepts may be related.

They are not necessarily identical.

The ability to transfer data without preserving those distinctions creates the appearance of interoperability without shared understanding.

Gobekli developed the idea of the Standardized Talent Asset Mapping Protocol, or STAMP, to explore how talent information could become more contextual, connected, and translatable across systems.

STAMP is not currently a ratified or independently governed standard. It is a developing conceptual and technical framework that continues to inform Gobekli’s Talent Tree and Human Intelligence Layer.

At its core, STAMP addresses one question:

How can systems connect different descriptions of human capability without erasing the meaning and context that make each description useful?

Data Portability Is Not Semantic Interoperability

Two systems are not semantically interoperable simply because they can exchange a file.

Several levels of interoperability must work together.

Technical interoperability

Can the systems connect and transmit information?

This includes APIs, authentication, transport protocols, and compatible infrastructure.

Syntactic interoperability

Do the systems recognize the format and structure of the data?

For example, can both systems parse the same JSON document and identify its fields?

Semantic interoperability

Do the systems understand what the data means?

If one system receives a field labeled “competency,” does it interpret that concept in the same way as the sending system?

Contextual interoperability

Do the systems understand how the information was created and how it should be used?

A skill inferred from résumé text, a capability demonstrated through a work sample, and a competency verified through an assessment may all use the same label. They do not carry the same meaning or level of trust.

Talent-data integration must address all four levels.

Otherwise, systems can exchange information while misunderstanding what it represents.

Why Flat Skills Tags Break Down

Skills tags are useful indexing tools.

They can improve search, organize content, and connect records with commonly used terminology.

The problem begins when a tag is treated as a complete representation of capability.

Consider the label project management.

A system cannot interpret that label responsibly without additional questions:

• What activity did the person perform?
• What responsibility did they personally hold?
• What was the project’s scale and complexity?
• What methods or tools did they use?
• What evidence supports the claim?
• What outcome resulted?
• When did this happen?
• Was the capability assessed, endorsed, inferred, or self-described?
• Which definition of project management is being used?

The same label could describe someone who completed an introductory course, coordinated a volunteer event, managed a complex technical implementation, or held a professional certification.

All four records may be valid.

They should not be treated as interchangeable.

A semantically interoperable system must preserve the relationships and distinctions surrounding the label.

Start With Experiences, Activities, and Evidence

STAMP proposes that talent data should be rooted in what people have learned and done.

An experience might include:

• A job
• A course
• A project
• An apprenticeship
• A credentialing program
• Military service
• Volunteer work
• Caregiving
• Creative practice
• Community leadership

Within an experience, a person performs activities, produces evidence, receives feedback, achieves outcomes, and develops or demonstrates capabilities.

This creates a connected structure:

Person → Experience → Role → Activity → Evidence → Outcome → Capability

Credentials, assessments, endorsements, and other records can verify or support particular parts of that structure.

The exact model may vary by use case. The important principle is that skills should remain connected to the experiences and evidence that give them meaning.

A Graph Is a Model, Not a Required Database

Talent information is naturally relational.

A person participates in multiple experiences. An activity may involve several skills. A work sample may support multiple claims. A credential may connect with a competency framework, occupation, learning opportunity, or licensing requirement.

Graph-based models are useful because they represent information as entities and relationships.

For example:

• A person performed an activity
• The activity occurred within an experience
• The activity produced an artifact
• The artifact supports a capability claim
• The capability aligns with a competency
• The competency belongs to a published framework
• A credential assesses or recognizes that competency

This graph can be implemented using different technical architectures.

A graph database may be appropriate for traversing complex relationships. A relational database can also represent those relationships through tables and keys. Document databases and hybrid systems may be appropriate in other environments.

STAMP should not require one storage technology.

The important requirement is that relationships, identifiers, sources, and meanings can be preserved and exchanged.

The Role of JSON-LD

JSON-LD is a W3C-recommended format for expressing Linked Data using JSON.

It allows familiar JSON data to be connected with globally identifiable concepts through a defined context. This can help systems understand that two fields with different local names refer to the same published concept—or that similar-looking fields refer to different concepts.

JSON-LD is useful for STAMP-like data because it can express:

• Identifiable entities
• Typed relationships
• References to external vocabularies
• Connections among records from different sources
• Data that can be interpreted as a graph
• A path from local application data toward web-scale interoperability

But JSON-LD does not make data semantically interoperable by itself.

Developers must still agree on:

• Which entities and relationships are being represented
• Which vocabularies define them
• How identifiers are assigned
• How provenance is preserved
• How mappings are created
• How uncertainty is communicated
• How access and privacy are governed

A technically valid JSON-LD document can still express an inaccurate or misleading relationship.

The format enables linked meaning. It does not guarantee the quality of that meaning.

Translation Without Forced Uniformity

Organizations have legitimate reasons for maintaining their own talent language.

An employer’s competency model may reflect its culture and workflows. A school’s learning outcomes may reflect a discipline. An industry framework may describe occupational requirements. A credential issuer may define specific assessment criteria.

Semantic interoperability should not require all of them to abandon those definitions.

STAMP explores a translation-oriented approach that Gobekli has called ontological refraction.

A local concept can be connected with concepts in other frameworks while retaining:

• Its original label
• Its original definition
• Its source
• Its relationship with the external concept
• The method used to create the mapping
• The degree of certainty involved

This matters because mappings are rarely all-or-nothing.

Two concepts might be:

• Equivalent for a defined purpose
• Closely related
• Broader or narrower
• Partially overlapping
• Prerequisites of one another
• Frequently associated but not semantically equivalent

A responsible crosswalk should preserve these distinctions.

Calling two concepts an “exact match” when they merely resemble one another can corrupt every analysis built on top of the mapping.

A Practical Translation Example

Imagine that an employer uses the internal capability:

Coordinate a cross-functional product launch.

A talent system might identify relationships with concepts such as:

• Project coordination
• Stakeholder communication
• Scheduling
• Risk management
• Change management
• Team facilitation
• Product operations

The original employer language should remain intact.

Each proposed relationship should also preserve its source and basis.

A structured record might show:

• The original capability and definition
• The activity through which it was demonstrated
• Evidence from the launch
• The person’s role and level of responsibility
• Relevant outcomes
• The external framework concepts considered related
• Whether each mapping was human-authored, imported, or AI-suggested
• Whether a reviewer confirmed the mapping
• Confidence or uncertainty where appropriate

Another organization could then interpret the record through its own framework without treating every related term as synonymous.

This is semantic interoperability as translation—not forced uniformity.

Provenance Is Part of Meaning

Talent systems frequently combine information from many sources.

Without provenance, the receiving system may not know whether a claim was:

• Entered by the individual
• Issued by a school
• Verified by a credentialing body
• Endorsed by a manager
• Extracted from a résumé
• Inferred by AI
• Demonstrated through evidence
• Imported from another platform

These distinctions are not secondary metadata.

They determine how the information should be interpreted.

A STAMP-informed model should preserve:

• Who created or issued the information
• When it was created
• What source record supports it
• Whether it has been modified
• How a semantic mapping was produced
• Whether a human reviewed an AI suggestion
• Which parts are verified and which remain claims
• What permissions govern its use

A sophisticated talent graph without provenance can produce sophisticated misinformation.

Time, Scale, and Context Must Travel With the Data

Human capability changes.

People practice skills, apply them in new contexts, combine them with other capabilities, and sometimes stop using them.

A talent-data model should therefore be able to retain relevant context such as:

Time

When did the experience occur? How long did it last? When was the capability most recently demonstrated?

Scale

What was the scope, complexity, consequence, or level of responsibility?

Environment

In what industry, organization, team, or operating context did the activity occur?

Role

Was the person learning, observing, contributing, leading, teaching, or accountable for the result?

Evidence

What supports the interpretation, and who can evaluate it?

Trust status

Was the information claimed, inferred, assessed, endorsed, or verified?

These dimensions should not automatically be collapsed into one proficiency score.

Different decision-makers may weigh them differently. The purpose is to preserve enough information for transparent interpretation.

How STAMP Relates to CTDL

The Credential Transparency Description Language provides a rich linked-data vocabulary for describing credentials, learning opportunities, assessments, organizations, costs, requirements, outcomes, pathways, and related resources.

CTDL-ASN supports the publication of competency frameworks and individual competencies, including relationships among them.

This is already sophisticated semantic infrastructure.

STAMP should not be positioned as filling an absence of meaning within CTDL or replacing CTDL’s role.

The distinction is primarily one of scope.

CTDL and CTDL-ASN describe ecosystem resources such as:

• What a credential is
• Who offers it
• What it requires
• Which competencies it addresses
• What learning opportunities relate to it
• Which occupations or pathways it supports

A STAMP-informed Human Intelligence Layer focuses more directly on how an individual person’s experiences, activities, evidence, and records connect with those resources.

For example:

• CTDL can describe a credential and its associated competencies.
• A verifiable credential can confirm that the individual earned it.
• TalentPass can connect that credential with the person’s other experiences.
• An activity record can show how related learning was later applied.
• Evidence or an endorsement can support the application.
• A TalentSync workflow can interpret selected information for a defined organizational purpose.

The opportunity is to connect these layers, not ask one standard to perform every function.

The Role of AI in Semantic Translation

AI can significantly reduce the manual effort required to structure and translate talent data.

It can help:

• Extract activities from natural-language descriptions
• Identify possible capabilities demonstrated through an experience
• Suggest relationships with external frameworks
• Compare differently worded competency definitions
• Detect overlapping or conflicting mappings
• Ask people for missing context
• Explain why two concepts may be related
• Prepare proposed mappings for human review

However, AI-generated mappings are interpretations.

They should not silently become facts.

A responsible system should retain:

• The original text
• The model-generated suggestion
• The source framework
• The reasoning or supporting evidence
• The confidence or uncertainty
• The identity of any human reviewer
• The accepted, corrected, or rejected result

This is particularly important when talent data influences hiring, advancement, education, licensing, or access to public services.

Semantic interoperability should make meaning more transparent, not hide consequential judgments inside an automated crosswalk.

From Semantic Interoperability to a Human Intelligence Layer

The Human Intelligence Layer depends on connections among people, experiences, capabilities, evidence, relationships, organizations, and intelligent systems.

If those connections are built from ambiguous labels, the resulting intelligence will be unreliable.

STAMP contributes a set of design principles for making the layer more meaningful:

• Anchor capability in experience and activity
• Preserve evidence and provenance
• Represent relationships explicitly
• Connect local terms with external frameworks
• Distinguish equivalence from similarity
• Preserve time, scale, role, and context
• Label AI-generated interpretations
• Keep the person involved in reviewing their own record
• Support exchange without requiring every system to use identical language

TalentPass applies these ideas to the individual’s living record.

The Talent Tree provides a developing structure for understanding experiences, capabilities, goals, and growth. Pythia helps people translate natural conversation into structured information they can review and use.

TalentSync applies related ideas to organizational opportunities, teams, learning, membership, and workforce workflows.

Together, they are intended to help individuals, organizations, and AI develop a shared understanding without pretending that every source describes people in the same way.

Interoperability Must Preserve Human Meaning

Talent-data systems do not need another universal list of skill labels.

They need reliable ways to connect different descriptions of human capability while preserving where each description came from, what it means, and how certain the connection is.

That requires more than APIs.

It requires:

• Shared identifiers
• Published vocabularies
• Typed relationships
• Provenance
• Context
• Responsible crosswalks
• Individual agency
• Transparent AI
• Governance appropriate to the decision

JSON-LD, CTDL, CTDL-ASN, LER standards, occupational frameworks, and verifiable credentials already provide important pieces of this infrastructure.

STAMP is Gobekli’s developing hypothesis about how those pieces can connect with the lived experiences and continuing records of individuals.

The goal is not perfect semantic uniformity.

Human capability is too contextual, dynamic, and culturally shaped for one universal vocabulary to capture completely.

The goal is responsible translatability: helping systems and people understand how different concepts relate without erasing their differences.

The future of talent mobility depends on more than moving data.

It depends on preserving meaning as the data moves.

---

STAMP status notice: STAMP is a Gobekli-developed conceptual and technical framework that continues to inform the Talent Tree and Human Intelligence Layer. It should not currently be interpreted as an independently governed, formally ratified, or broadly adopted public standard. Its data models, terminology, and implementation may continue to evolve.

Product availability notice: This article discusses Gobekli’s broader technical vision for semantic interoperability and the Human Intelligence Layer. Some capabilities described may be in development, expanding release, or available only through Launch Partnerships. Visit the Product Availability page for the current status of TalentPass, TalentSync, Pythia, Profiles, Passport Pages, and the Talent Tree.