mbrush
commented
4 years ago The Agent model proposed for VA is minimal - providing only a few generic attributes common to all types of agents. Future versions may expand on this specification, but at present it is up to implementations to extend this base model as needed to meet their use cases.
Here we define an abstract Agent class, with Person and Organization as two concrete specializations that inherit all attributes of Agent, but do not define any unique attributes of their own. 'Computational' agents pose nuanced challenges - related to the grey area / duality of software as an autonomous actor (agent) and software as an algorithmic specification (method). As they are not a high priority use case - we defer modeling these for now.
Modeling Proposals
Based on modeling approach to Agents in the CD2H Contributor Attribution Model defined here.
Agent
- id: identifier (1..1) # A unique string that identifies the agent.
- type: class (1..1) # The high-level class to which the agent belongs (‘Person’, ‘Organization’, or ‘Computational Agent’).
- label: string (0..1) # A free-text name for the agent.
- description: string (0..1) # A free-text description of the agent.
- externalID: identifier (0..1) # Additional identifier(s) for the agent that come from an external system or authority.
- url: url (0..m) # A URL where information about the Agent can be found.
Proposal also captured in the modeling spreadsheet here.
Person
- Persons are human beings with agency who contribute to artifacts. Person is a concrete sublass of Agent. It can be instantiated in the data, and inherits from its abstract parent Agent class. At present the we do not define additional person-specific attributes - preferring a minimal but extensible model that allows implementations control over how Persons are represented.
Organization
- Organizations are groups of agents (typically persons) structured and managed to achieve a common goal. Like Person, Organization is a concrete subclass of Agent. It can be instantiated in the data, and inherits from its abstract parent Agent class. But at present the we do not define additional organization-specific attributes.
Agents are autonomous actors that bear some form of responsibility for an activity taking place, for the existence of an entity, or for another agent’s activity.
An agent can be an individual Person, an Organization of multiple individuals acting together, or a Computational Agent such as a software program or algorithm.
In the VA model, representation of agents is needed to capture who interprets/annotates variants, who generates data used as evidence, who submits annotation records to knowledgebases, and who reviews and approve submitted annotations.
Person agents can belong to organizations, and act on behalf or in the context of an Organization when contributing to an Artifact. For example, curators generating these variant pathogenicity interpretations do so in the context of the ClinGen organization, which provides administrative, technical, and funding support for their contributions.
A related ticket on modeling Contributions is here.