Bdi Normative Reasoning

---
name: bdi-normative-reasoning
description: Design an agent's handling of obligations, prohibitions, permissions, and conflicts among norms and goals. Use when an autonomous agent must decide which norms apply and what to do when they conflict. NOT for simple fixed policy checks, ordinary BDI commitment design, or organizational metaphor.
license: Apache-2.0
metadata:
  provenance:
    kind: semantic-merge
    sources: [normative-bdi-agent-architecture, normative-bdi-agents, a-normative-extension-for-the-bdi-agent]
---
# BDI Normative Reasoning

Use this skill when norms are represented as decision-relevant objects, rather than a fixed `if` statement at an external enforcement boundary.

## Decision procedure

1. Specify each norm's issuer, scope, activation condition, addressee, required or forbidden action, expiry, and enforcement consequence. Keep a recognized Abstract Norm Base (ANB) candidate distinct from a grounded active Norm Instance Base (NIB) instance.
2. Ground applicability in source-tagged current beliefs; distinguish “not observed” from “known false.” Instantiate only after confirming binding, activation, and non-expiry under a declared reconciliation policy.
3. Keep norm recognition separate from adoption as a goal or intention. An applicable norm need not be internalized if the design permits refusal.
4. Detect conflicts among norms and with existing intentions using a joint plan witness. Separate an empty complete plan catalog from incomplete search or unknown effects. Name the chosen resolution criterion: an applicable hard priority, a scoped partial-order comparison, or explicit escalation. Preserve incomparable alternatives.
5. Produce a decision trace showing applicable norms, rejected alternatives, source authority, search coverage, and the rule used. Keep any local desire update separate from authorization to execute an external effect.

```mermaid
flowchart LR
  B[Beliefs] --> G[Ground norms]
  N[Norm catalog] --> G
  G --> C{Conflict?}
  C -->|no| A[Adopt or comply]
  C -->|yes| R[Resolve by declared policy]
  R --> A
  A --> T[Decision trace]
```

## Boundaries

- Do not use consequence ranking as a universal conflict resolver. The cited model can leave consequences incomparable; a tie policy must be separately declared. Verified hard constraints still bound feasible actions.
- External policy enforcement may be simpler and safer than autonomous norm adoption. Choose this architecture only if the agent truly has discretion.
- `bdi-agent-architecture` handles basic goals and commitments; `bdi-agent-interpreters` handles the executable cycle.

## Source-bound checks

- Read `sources/normative-bdi-agent-architecture/references/source-boundary-and-authority-trace.md` for primary-source scope, ANB/NIB distinctions, and independent effect admission.
- Read `sources/normative-bdi-agents/references/source-boundary-and-authority-trace.md` when comparing source variants. The similarly named 2014, 2015, and 2012 works are distinct; do not collapse their results.
- An active NIB instance is a local model state, not a legal, ethical, or execution authority.

## Source bundles

The three imported variants are preserved under `sources/normative-bdi-agent-architecture/`, `sources/normative-bdi-agents/`, and `sources/a-normative-extension-for-the-bdi-agent/`. Compare them before quoting a paper-specific rule; verify claims against the primary publication.