Skip to content

6. Obligations, Permission & Power

Obligations, Permission, and Power

Institutional constraints on action in AgencyGrid can be expressed as three main elements:

  • Obligations – A duty for an agent to bring about a specific institutional outcome. Once the outcome occurs, the obligation is fulfilled.
  • Example: In a distributed problem-solving setting, a coding agent is obligated to deliver a software module after receiving a specification. The obligation ends when the module is submitted.

  • Some obligations depend on external confirmation, such as an evaluator verifying correctness and compliance with SLAs.

  • Permissions – An indication of whether an action is allowed in a given institutional state.

  • If an action is not permitted and is still carried out, it violates institutional rules.

  • Some violations are serious (e.g., closing submissions before the deadline), while others may be minor and ignored.

  • Powers – The ability to take an institutional action that changes the institutional state.

  • Powers only apply to actions recognized by the institution, not to external-world actions.
  • Example: In the work submission scenario, an evaluator has the power to open and close submissions, while a coding agent does not.
  • Powers can be granted or removed based on institutional processes.

In all three cases, the action must first be recognized as relevant to the institution in the current state. Only then can it be checked for permission, obligation, or power, and any resulting institutional changes can take place.

Evolution & Adaptation

Agency evolution is the ongoing process of adjusting to new requirements and environmental changes.

The main purpose of an agency is often to maintain stability — yet both agencies and environments inevitably change. Agents may join or leave, goals can shift, and environmental conditions can evolve, requiring agencies to adapt.

Reorganization can happen for two main reasons: as a reaction to local changes in the environment, or as a way to implement new overall goals or strategies. Therefore AgencyGrid not only individual agents to adapt but also the agency's structure to change dynamically when needed. Depending on the situation, adaptation can be done through changes in agent behavior, updates to interaction agreements, or a complete re-design of the social structure.

The challenge is to know when it is better to reorganize, even if it temporarily reduces stability, and when it is better to maintain things as they are. This decision depends on the agency's utility, which is the degree to which it meets its objectives. Reorganization is worth doing if it increases utility, meaning the agency performs better afterward. However, utility also depends on the cost of making changes, so both the benefits and costs must be considered.

An agency's performance is influenced by its structure, the way tasks are arranged, and how decision-making is organized. In changing environments, the ability to adapt these elements is key to maintaining or improving performance. Hence agencies that adapt their structure to match environmental changes tend to perform better.

Dynamic adaptation is the ability to change the structure or behavior of a system while it is running, without shutting it down. This can include adding, removing, or replacing components. For this to work, systems need to be able to evaluate their own health, metrics and take actions to protect or restore it. In agencies, this means reorganization can involve changes in how agents operate, such as adding or removing members, or changes in the social structure, such as roles, relationships, norms, or interactions.

Emergence

AgencyGrid supports both top-down and bottom-up organization design, and can blend them into hybrid emergent structures:

  • Top-Down Design
  • External authorities or designated decision-makers define structure, goals, roles, and operational rules.

  • Best suited for stable environments with well-understood objectives, clear stakeholder requirements and stability is prioritized over adaptability.

  • Bottom-Up / Self-Organization

  • Agents coordinate and adapt based on local interactions, shared incentives, and environmental feedback.
  • Effective in dynamic, complex environments where no single actor has complete knowledge.

  • Novel solutions emerge through distributed decision-making.

  • Hybrid Emergent Structures

  • Combine high-level objectives and guardrails from top-down design.
  • While allowing the specifics of coordination, resource allocation, and problem-solving to evolve bottom-up.

Emergence in AgencyGrid is enabled by:
- Feedback loops.
- Adaptive role reassignment.
- Swarm intelligence principles.
- Evolving incentive structures.

The result is organizations that self-restructure in response to shifting goals, new participants, and unexpected challenges, balancing stability with adaptive capacity.