Hyperlocal dashboard IA Stage 1 — five locked architectural decisions
Attribution
Prior state
No documented information architecture for the local-hubs
dashboard. Open architectural questions: aggregation invariant
across tiers, trust-ring model, GIS substrate, FCI behaviour
across scales, federated-AI status visibility.
Change
Five architectural decisions locked for Stage 1:
(1) Matryoshka is one schema, five granularities — same four
pillars × five nested scales. Every Community indicator must
declare its upstream join into City. Aggregation invariant
is the load-bearing rule.
(2) Hubs produce at Community, subscribe everywhere else —
Planet/Bioregion subscribed (foundation models), Region/City
harvested (open data portals), Community produced (hub
sensor campaigns + GIS + SLMs). Community tier is the only
one that requires PLANETAI's presence.
(3) Three concentric trust rings on one surface — public
(resident) / council (attested) / hub-ops (credentialed).
Same surface, gradual disclosure on attestation. Maps onto
PKC sovereignty-by-architecture.
(4) Hyperlocal GIS as editable substrate, not basemap — hubs ADD
layers, council publishes/restricts, residents read what's
permitted. CENTRINNO Cartography blueprint is the
methodological substrate.
(5) FCI is matryoshka itself — five FCIs, one per tier, computed
two ways: synthetic-from-above (subscribed/harvested data)
and aggregated-from-below (rolled up from Community).
Divergence Δ between the two is itself a first-class
indicator. Boundary definitions of PITO are tier-specific —
at Planet, PITO is redefined as planetary-boundary
disturbance rate (Richardson 2023). ΔFCI/Δt also matryoshka.
Five Tomas-resolved design questions (4 May):
- Trust rings confirmed (gradual disclosure; revert to tabs only
if mockups fail)
- FCI varies across scales (each tier shows its own FCI
prominently; vertical FCI strip is the comparison view)
- Customary-authority sign-off as Governance indicator only,
for now (Bali Tri Hita Karana edge case deferred)
- Tier names localized per place, fractal approach (universal
schema T0–T4 underneath, place-specific labels on top)
- Federated-AI status visible, not hidden (sensor mesh, agent
runs, model weight sync, embedding tier surfaced per-ring)
Reasoning
IA decisions need to survive into Stage 2 (mockups) and Stage 3
(data hookup). The matryoshka aggregation invariant and the
trust-ring model are easy to silently break in mockup work if
not held explicitly. The FCI dual-aggregation (synthetic vs
aggregated, divergence as deliverable) is the methodological
contribution PLANETAI offers beyond Boeing 2024 — losing it in
UI translation would forfeit a load-bearing differentiator.
Artifacts changed
- Google AI and Science/PLANETAI_hyperlocal_dashboard_IA_2026-05-04.md
Review
Required acks (0/3 acks): cast.foundation, mdg.fab-lab-bali, it-del.ben-koo
Window closes: 2026-05-25
anchor pilot for Stage 2 mockups = Bali (per track-bali-hyperlocal-dashboard)
Related decisions
Edit this decision on GitHub →
New here? See how to contribute →