Research

Evidence Backbone · Candidate v0.9.0-rc.1

The CDI-BoK does not use “Deep Research” as a synonym for validated evidence. AI-assisted syntheses can locate concepts, tensions and candidate references; every material source must be checked against its original and assessed against the claim it is intended to support.

Start with your question

If you need to…	Route	Outcome
understand DI and the maturity of the field	DI state of the art	foundations, boundaries, disputes and limits
know what authority each PULSE component has	PULSE evidence map	constitutional, derived, candidate, hypothesis or unevaluated
use trends without treating them as facts	Future signals	scenarios, horizons and update conditions
decide what to study next	Research agenda	falsifiable questions and minimum designs

How knowledge is organized

Layer	Examples	Rule
Established foundations	Decision Analysis, DSS, information quality, outcome bias	primary or official source and application boundaries
Mature evolving practices	DMN, EDM, data governance and AI risk	version and as_of date
Emerging fields	human–AI deliberation, conversational analytics and agents	context-dependent evidence
PULSE synthesis	PDAMR, Decision Circle, five verbs and Human-in-Control	project authority, not independent effectiveness
Hypotheses and metaphors	allostasis, interoception and decision organism	observable prediction and abandonment possibility
Future signals	expected changes in interfaces, agents and governance	scenario, not fact or readiness

Verifiable backbone

The governance/registries/evidence-profiles.yml registry preserves hashes, provenance, completeness, rights, permitted use and eight dimensions without turning them into one total score. ADR-028 blocks incomplete sources, AI-assisted syntheses as direct support and forecasts used as facts.

Central risk: confusing novelty of language with theoretical novelty, and theoretical novelty with demonstrated effectiveness.