Knowledge Impact Framework

How Domain, Pillars, Action Kinds, Actions, and Evidence tiers pull together into a coherent causal accountability structure — and how the Biosphere → Society → Intervention chain binds them.

Domain

Plant-Based Food Transition

"Industrial animal agriculture is a primary driver of biosphere destabilisation; transitioning to plant-based food systems simultaneously addresses climate stability, land use, biodiversity loss, and equitable access to nutrition."

onto_chain — the causal premise injected into every narrative prompt

defines 4 pillars
Pillars — argument axesthe WHY — what societal stakes are at risk
Food Security

Biosphere harm: Land routed through livestock at 10–15% caloric efficiency

land_useprocurementfinancialgovernment_motion
Climate Resilience

Biosphere harm: Methane, nitrous oxide, and land-clearing from animal farming

climatebiodiversityland_usetreatylegislation
Equity & Access

Biosphere harm: Land concentration, food cost inflation, diet-related disease burden

cateringsocial_carefiscal_policyprocurement
Biodiversity & Nature Recovery

Biosphere harm: Deforestation, soy monocultures, and pasture conversion

biodiversityland_useplanningsubsidy_design

Each pillar = a distinct societal stake + a specific biosphere harm that animal agriculture causes. One action can serve multiple pillars simultaneously.

each action tagged to kinds + pillars
Action Kinds — 18 policy instrument typesthe HOW — what type of intervention
cateringprocurementland_usebiodiversityclimateeducation
financialfiscal_policyregulatorysocial_carelegislationplanning
statutory_instrumentsubsidy_designgovernment_motionprocurement_mandateinternationaltreaty

Each Action Kind can serve multiple pillars simultaneously — e.g. land_use actions appear under Food Security, Climate Resilience, and Biodiversity.

kinds group into
Actions — specific trackable commitmentsthe WHAT — the actual policy

Plant-based default in council catering

cateringequity

Rewild former grazing land

land_usefood_security + biodiversity

Meat/dairy consumption tax

fiscal_policyequity + climate

Hospital plant-based meal share

procurementequity + food_security

Each action carries: action_kind (instrument) · pillar[] (which argument axes it serves) · applicable_tiers (council / national / parliament / devolved) · why_rationale (grounding text for LLM prompts)

actions draw evidence from
Evidence Library — three retrieval tiersthe BECAUSE — grounded in peer-reviewed research
Tier 1cross_domain

Universal Premises

Applies to every domain, every pillar

  • · IPCC Climate Change 2022
  • · Safe and just Earth system boundaries (Rockström)
  • · SDG framework

domain_tag: null → matches any domain

pillar_tags: empty → matches any pillar

quality_score: 0.55–0.95

Tier 2domain

Domain Foundation

Applies to all actions within the domain

  • · Poore & Nemecek 2018
  • · UK Food Security Report 2024
  • · National Food Strategy

domain_tag: food_transition

pillar_tags: empty → matches any pillar in domain

quality_score: 0.55–0.95

Tier 3action_specific

Policy Instrument Evidence

Supports specific action kinds via paper_action_links

  • · Food tax modelling study
  • · School meals RCT
  • · Procurement impact analysis

domain_tag: specific domain_tag

pillar_tags: specific pillar tags set

quality_score: 0.30–0.85

Retrieval uses pgvector cosine similarity weighted by quality_score (meta-analyses outrank grey literature at equal semantic distance). Tier 1+2 papers feed pillar narratives; Tier 3 papers feed action-specific summaries via paper_action_links.

The Causal Reasoning Chain — embedded in every generated narrative

Biosphere → Society → Interventionthe reasoning chain embedded in every narrative prompt

Layer 1 — Biosphere

What planetary/ecosystem condition makes this action urgent? Grounded in cross_domain + domain papers. Specific mechanisms and quantities.

"Animal agriculture routes 80% of agricultural land through livestock at 10–15% caloric efficiency — land that could feed 5–10× more people directly."

Layer 2 — Society

How does that biosphere condition specifically threaten the domain's human stakes? UK context, societal consequences.

"UK food self-sufficiency falls; import dependency rises; diet-related disease burden concentrates in lower-income households who cannot access affordable protein alternatives."

Layer 3 — Intervention

How does acting on this pillar/action address the causal chain? What leverage does the evidence show?

"Shifting school and hospital catering to plant-based defaults directly reduces institutional demand, demonstrates price parity, and normalises dietary change — evidence shows 30–60% uptake when plant-based is the default."

This chain is directional: ecological conditions first → human stakes → policy leverage. The onto_chain from the domain is injected as grounding context in every call, ensuring all narratives inherit the same causal position.

produces
Generated Outputs

/why

Domain-level pillar narratives from Tier 1+2 papers. 4 pillars × N papers cited.

/actions/N — Foundation

Per-action pillar cards from Tier 1+2 RAG. One card per pillar × papers cited.

/actions/N — Evidence

Action-specific supporting/contesting summaries from Tier 3 papers via paper_action_links.

/council/N — Causal Impact

Per-decision causal chain: what this vote enables or forecloses, cited to evidence.

Key Structural Distinctions

Pillars ≠ Kinds

catering is a Kind (instrument). equity is a Pillar (argument). A catering action can serve equity, food security, and climate simultaneously. The separation prevents circular reasoning — you can't justify a policy instrument by the same instrument.

Tiers are retrieval scopes, not quality ranks

A Tier 3 paper (action-specific RCT) may have a higher quality score than a Tier 1 report. The tier determines which RAG query path retrieves it — foundation queries vs. action-link queries — not its evidentiary weight within its scope.

onto_chain is the binding glue

Without the domain's causal premise injected into every prompt, pillar narratives would drift toward generic climate or health messaging. The onto_chain keeps all generated text anchored to the specific claim: because industrial animal agriculture does X, acting on pillar Y achieves Z.