Agent
Anything that can perform an action. Has agency, i.e. can also plan to take actions.
Agents can be individuals, groups, projects, networks. Agents can also be bots, machines (IoT)
Related to REA / NRP-CAS / NDO
Agent / Affiliate vs User
In a p2p context it is not advised to reduce agents or affiliates to users.The term user is appropriate for centralised platforms (Facebook, Uber) where participants have no control over the platform, interact on terms set by the owner, and provide data in exchange for access. An agent, by contrast, is an entity with agency, i.e. capacity to plan and take actions, to own its own data, and to participate in defining the rules of the system it operates within.
This is not merely semantic. The data model follows the ontology: a "user" has an account on a server that another party controls. On p2p infrastruture (ex. Holochain), an agent has a cryptographic key pair that they control, a source chain of their own actions that only they can append to, and relationships with other agents that are maintained on a shared DHT, not a shared database controlled by an operator. Holochain is the first infrastructure natively designed for agent-centric data models.
In the old REA / NRP-CAS language we have users and economic agents.
A user is linked to an account, it is credentials to access the NRP-CAS.
An economic agent can represent an individual, a project / venture, another network or even a traditional organization. Economic agents (or simply agents) ca be of various types: agent type.
Agent that contribute to processes are affiliates. Furthermore, affiliates have profiles, can form a digital identity, and that in the case of a NRP-CAS as a platform (server-client architecture) user is the credentials to access the platform.
We can say that resources are used in events by agents (hence REA).
User vs agent
The term user is used in the context of online centralized platforms such as Facebook or Twitter. The users in these contexts have no control over the platform, they are provided the platform as a service, in exchange of user data produced through interactions within the platform, which can be monetized elsewhere by the platform owners.
User also has a local connotation, as an agent within a specific application. Agent is used in a more generic sense that can transcend the local contexts.
In a p2p context it is not advised to reduce agents or affiliates to users.
The Agent Type Spectrum
In Valueflows language, an agent is anything that can perform an action. Therefore an agent has agency, i.e. can also plan to take actions.
In a P2P complexity economics framework, the system must model:
- Individual agents: human participants, each with a source chain
- Collective agents: groups, working groups, communities — entities with collective agency
- Project/venture agents: temporary organised efforts with their own goals and resources
- Network agents: larger-scale communities (an OVN as a whole is also an agent in its ecosystem)
- Artificial agents: bots, AI models, IoT devices — increasingly significant actors in the commons
Individual vs Person
The two layers of agent identity. This distinction is of deep philosophical and technical importance: An individual is the physical you, you exist and therefore you have access to certain things. A person is a representation of the physical you, it is an agent in a socio/economic context.
In practical terms:
- Individual-type identity: exists by virtue of holding a cryptographic key pair. Permissionless access. No disclosure required. Like Bitcoin: you only need keys to transact. This is the foundation of the NDO's permissionless entry model.
- Person-type identity: requires reputation, credentials, or demonstrated capabilities earned through participation. Required for sensitive processes: governance participation, custodianship, specialised service provision.
This two-layer model is not bureaucratic gatekeeping, it is a precise answer to the information-theoretic question: "what does this process require to function well?" Taking the bus requires a ticket, not identification. Performing surgery requires credentials. The NDO must distinguish which processes are ticket-level (individual-type) and which require person-type identity, and build the appropriate mechanisms for each.
In complexity economics terms: forcing all interactions through person-type identity is a high information-overhead solution applied to low-information-need situations. It reduces participation, increases friction, and creates centralisation risks (who issues the credentials?). The OVN model, and therefore the NDO, reserves person-type identity for the specific processes that genuinely need it.
Identity ≠ Credentials
When you take the bus you don't need to tell the driver all about yourself before boarding, you just show your ticket, which is the credential required to access the transportation service.
Credentials are context-specific access tokens derived from past actions or present conditions. They are not identity. You can prove you can safely use a 3D printer (credential) without revealing who you are (identity). This distinction, made precise through zero-knowledge proof technology, is the foundation for privacy-preserving participation in P2P systems.
The implications for NDO agent design: the system should be designed so that agents prove what they can do through minimally-revealing credentials, not by revealing who they are through full identity disclosure. This is currently only partially implemented: the PPR system generates verifiable contribution records (a form of capability credential), but there is no ZKP layer to share them without revealing the underlying data.
Agent as Network Node: Social Capital and Interconnection
The individual profile description captures a dimension of agents often overlooked in technical implementations: an agent's profile includes "social relations (connections with other agents), which can be seen as social capital, may be shared in context / processes, part of the wealth structure of a network" and "network affiliations."
Social capital: the value embedded in relationships and network position, is an intangible resource that the NDO system should make legible, not obscure. In complexity economics, information flows through social networks, and an agent's connectivity is a measure of their access to collective intelligence. The individual profile should capture relational wealth, not just transactional history.
The Long Tail of Agent Engagement
From the governance archive, the OVN operates with a 1-9-90 engagement distribution: 1% core (entrepreneurial, high commitment), 9% active contributors, 90% occasional participants. This is not a design flaw, it is universal in open networks. The agent model must be designed for this reality.
An agent system designed only for the 1% (assuming high engagement, rich profiles, continuous participation) will exclude the 99% who should still be able to participate meaningfully at their actual engagement level. The affiliation model (unaffiliated → close → active → core) is the OVN's answer to this: design for the full spectrum, not just the core.
The affiliation spectrum
The OVN model uses five affiliation states that capture the long tail of participation:
- Unaffiliated: anonymous; accesses public information only; no obligations; not known by affiliates.
- Close affiliate: not yet contributed in a trackable way; known by some affiliates; can access database and some activities; no formal obligations but respects OVN's assets.
- Active affiliate: has logged contributions in the NRP-CAS AND is recognised by the community; has access to OVN infrastructure, governance, and production assets; must adhere to the Benefit Redistribution Algorithm and affiliation conditions.
- Core affiliate: algorithmically selected from active affiliates based on contribution + reputation; special privileges within specific projects or ventures; status is dynamic (in/out based on contribution; can expire through inactivity).
- Inactive affiliate: previously contributed, currently inactive; access to some activities; can re-engage by contributing again.
Formal affiliation: becoming an active affiliate formally consists of:
- Acknowledging the Affiliate Handbook
- Acknowledging the Nondominium & Custodian agreement
- Signing the "Acknowledgement of Knowledge Commons" agreement
- Signing the Benefit Redistribution Algorithm agreement
This allows the system to algorithmically determine who should participate in a governance decision (active affiliates only? core affiliates? anyone with a Person entry?), to manage the onboarding process, and track affiliation lifecycle.
Affiliation state as a derived property
In NDO, rather than storing affiliation state (which would need maintenance and could become inconsistent), derive it algorithmically from existing DHT data:
affiliation_state(agent) = f( person_exists(agent), // Boolean contributions_count(agent), // From economic events last_contribution_timestamp(agent), // Recency from economic events reputation_summary.total_claims, // From PPRs affiliation_record_exists(agent), // Has the agent signed the ToP? governance_claims_count(agent) // From PPRs )
→ Unaffiliated | CloseAffiliate | ActiveAffiliate(contribution_count) | CoreAffiliate(governance_weight) | InactiveAffiliate
CoreAffiliate is algorithmically determined: an active affiliate whose PPR (private participation receipts) contribution rate in the past N days exceeds a configurable threshold becomes a core affiliate. This implements the OVN governance equation: governance access as benefit derived from contribution activity.
Complexity economics justification: why each agent concept matters
Agent type taxonomy matters because economic models must faithfully represent economic actors. In Sensorica's actual operations, the entities performing actions are networks, ventures, and partner organisations, not just individuals. Flattening these to individual-human agents produces systematically inaccurate economic accounting. Benkler's P2P advantage, that P2P systems can process more information and model more finely-grained reality, is lost when the agent model is coarser than the actual social structure.
The affiliation spectrum matters because governance decisions made by binary "in/out" membership are weaker than those made by graded participation-aware membership. The governance equation (contribution history → governance access) requires a nuanced engagement model as input. If everyone is either "in" or "out," the governance equation reduces to "everyone who joined gets an equal vote," which ignores the 1-9-90 engagement reality. The result is either oligarchy (the 1% who do all the work decide everything while the 90% technically could override them) or paralysis (decisions require the majority of a population that mostly isn't paying attention).
ZKP capability proofs matter because the current privacy-accountability model forces a false binary: either share raw data (low privacy, high accountability) or share nothing (high privacy, no accountability). Zero-knowledge proofs break this binary. An agent can prove "I have at least 10 completed maintenance commitments" without revealing who the counterparties were, what the resources were, or what the exact performance scores were. This is the technical foundation for privacy-preserving meritocracy, governance access based on contribution without requiring surveillance.
Credential portability matters because the value of contribution data is severely limited if it only functions within a single application instance. An experienced Sensorica contributor who joins a new OVN should be able to leverage their track record. Without portability, every new commons requires starting from zero, creating massive friction to growth of the P2P ecosystem. Benkler's model of the peer production advantage assumes information flows across organisational boundaries, credential portability is the mechanism that makes this flow possible.
The composable individual profile matters because an agent's full profile, their roles, contributions, social relations, affiliations, capabilities, is the information substrate for most governance and economic decisions in the commons. If this information is scattered across disconnected entries with no composable view, the information cost of making well-informed decisions is high. Bar-Yam's complexity matching principle: the governance system's access to information about participants must match the complexity of the participation patterns it is trying to govern.
Collective agent types matter because commons-based peer production is fundamentally about collectives. Modelling it as a collection of individual agents without collective agency misses the emergent properties of collaboration. A working group is not just the sum of its members, it has its own identity, its own resource access, its own reputation, and its own governance. Without collective agent types, the NDO cannot model the holonic structure of OVNs: the fact that every entity is simultaneously a whole and a part of a larger whole.
