GQ-010 — How do we increase the People Graph's value by 100×?

How do we increase the People Graph's research and decision value by 100× without treating row count as value or weakening identity quality, provenance, rights, privacy, reproducibility, and reversibility?

SISO Open Source Foundation

active

scoped

Not declared

Define the minimum public-safe question program before scheduling research.

gls:release:f8ddbc97-e795-474a-b9c1-e508bd9e1787 · program-metadata-1.0.0+57320d3

Not released — the selected Release contains question-program metadata, not an answer artifact.

hybrid

Choose which source classes, identity methods, temporal and claim layers, query surfaces, and operating loops to fund, merge, revise, or kill for the next evidence-backed People Graph release.

Breadth: demonstrate at least tenfold source-native observation coverage across four distinct source classes, with every promoted observation carrying source, native identifier, observed-at time, rights state, provenance, and receipt references. · Resolution quality: on a versioned stratified benchmark, auto-accepted identity links achieve at least 99.5% precision, recall and abstention are reported separately, and every accepted merge or alias remains reversible with literal evidence. · Research depth: for at least 100 prioritized people or 25 named research questions, expose a three-layer path from raw observation to cited claim to explicit adjudication, with no unsupported cross-source assertion presented as fact. · Decision value: in at least five pre-registered operator or research decisions, the graph changes the selected action or reduces evidence-retrieval time by at least 80% against a measured baseline; additional rows alone cannot satisfy this criterion. · Rights and privacy: every promoted observation has an explicit rights state, restricted or discovery-only material never appears as a public payload, and living-person data is limited to justified public or authorized evidence. · Reproducibility and cost: the accepted increment rebuilds from versioned manifests, produces comparable quality reports, and stays within a declared per-source acquisition and refresh budget.

Source expansion increases rows or edges by at least tenfold but produces no measured improvement in the five decision-use cases. · Auto-accepted identity precision falls below 99.5%, accepted links cannot be reversed, or ambiguous names are silently merged. · Any promoted public record lacks a rights state, leaks restricted payloads, or exposes unjustified sensitive information about a living person. · New source pilots mostly duplicate existing low-signal entities and cannot produce replayable source-native observations with receipts. · The graph remains a collection of names and topics without cited claims, temporal context, contradiction handling, or adjudication. · Refresh and build costs exceed the declared budget without a proportional improvement in resolution, research depth, or decision value.

There is no agreed baseline for current evidence-retrieval time, decision changes, or operator utility. · The exact People Graph source commit has not been proven as a clean reproducible build in this registry lane, and the reviewed v2 builder's schema path does not match the tracked schema location. · A versioned identity-resolution benchmark with precision, recall, abstention, collision, living-person, and historical-person strata is not yet published. · Source-specific terms, privacy, retention, and redistribution rules are incomplete for the proposed scholarly, software, AI, creator, and media pilots. · The pg-observation-0.1 interchange has not yet been exercised end to end through Foundry discovery, Evidence Engine adjudication, and canonical graph admission. · Historical references under a different GitHub owner remain a verification and migration question rather than an assumed alias.

Any identity benchmark regression, disputed auto-merge, rollback, or evidence-loss incident. · Any source terms, rights, privacy, API, rate-limit, or retention change that alters admissible observations. · A parallel source pilot publishes a replayable observation set and draft handoff. · The v3 schema, identity, query, claims, build, or integration lane publishes a compatibility contract. · The first five decision-use cases receive measured baselines and post-change outcomes. · Review at least monthly until a first Answer Release, and immediately after a material source, schema, quality, or governance change.

SISO People Graph code and contract at de048bb3b34bf931b56fd741cb46c1334acdfb98 · SISO Book Library code and contract at be9ab0831b9ea8802d6898f3a3dfa8a61c63e80b · Publication-safe outputs from the thirteen named parallel lanes and their draft handoffs · Public or authorized source observations expressed through pg-observation-0.1; raw restricted payloads remain outside the registry

A source-by-source and capability-by-capability decision matrix separating breadth, resolution quality, research depth, decision value, rights and privacy, reproducibility, and cost, with explicit fund, merge, revise, or kill recommendations and traceable evidence.

Reassess on every material lane handoff and at least monthly until a first evidenced Answer Release; afterward refresh on source-contract, rights, schema, identity-quality, cost, or decision-utility triggers.

public metadata only