Spoke

reincarnation

Adaptive psychometric diagnostics — , , live .

Character

A people analytics professional running engagement / pulse / construct-validity work. You know what a Cronbach α is and you've never seen one tracked live in any HR platform you've used.

Problem

External. Your pulse surveys are static. Items don't retire when they stop discriminating. Reliability is computed once a year if at all. New items get added by survey design committee, not by information gain. Your respondents see the same questionnaire month after month even though half the items have asymptoted.

Internal. You're seen as "the Likert mean person" when you'd rather be the psychometrician. Your stats friends in marketing or product run real IRT models; you run pivots in Sheets.

Philosophical. IRT and adaptive item selection have been solved in education and clinical assessment for decades. Why does no HR platform ship them?

Guide

Reincarnation watches which survey questions are actually telling you something and which ones aren't, item by item, in real time. Underneath, it uses Item Response Theory — the same framework academic psychometricians use — and tracks a reliability coefficient called Cronbach α as responses come in. reincarnation — the toolbox's adaptive psychometric diagnostic engine. Information-gain-weighted item selection. Six-stage pool lifecycle (D / C / B / A / PURGATORY / INFINITY). Cronbach α tracked live per construct. Universal items (RIDs) shared across studies; study items (SIDs) carrying virtue + IRT + reliability.

Abstract

Background. Workforce surveys overwhelmingly field static forms; latent-trait estimation and adaptive testing are textbook in assessment science but practically absent from HR tooling, leaving analysts with stagnant item banks and unexplained shifts in pooled means.

Methodology. The spoke implements galaxy-scoped latent measurement: universal items plus study-scoped variants with virtues and three-parameter logistic–style discrimination parameters. The engine assigns items to graded pools (D→C→B→A with escalations toward INFINITY), performs information-gain–weighted adaptive selection honoring pool constraints (PURGATORY retired items withheld from selection), computes live Cronbach α on ingested respondent vectors, and exposes a construct-level psychometric feed for dashboards or downstream cards.

Scope. Provides selection, ingestion, persistence, and read APIs—not a SaaS survey UI. Does not adjudicate fairness or causal claims on its own; pairs with calculus for aggregates and segmentation-studio or pay-fairness for cohort linkage when cross-group reporting is needed.

Contribution. Responses and contracts are Zod-typed across HTTP and MCP. Selection and ingestion are service-key gated (POST discipline per PAT-11); read feeds remain public-inspectable. Galaxy + item payloads are idempotent, enabling consumer repos to register constructs at deploy-time without brittle SQL seeds.

Evidence / Provenance. Greenfield toolbox implementation on schema reincarnation with Performix-aligned seed galaxies; methodological lineage follows standard IRT and adaptive testing literature summarized in spoke README.

Plan

01
Register or extend galaxies
POST /api/spokes/reincarnation/galaxies registers constructs, RIDs, and SIDS idempotently; POST /api/spokes/reincarnation/galaxies/[id]/items appends rows later without hand-edited SQL.
02
Reincarnation handles selection
Ask for next items via reincarnation.adaptive-selection; the engine returns the highest-information items respecting active pools (PURGATORY / INFINITY semantics).
03
Ingest responses
Bulk-post via reincarnation.responses-ingest. Reliability is recomputed live; item statistics update; the pool lifecycle advances.
04
Read live psychometric health
Hit reincarnation.psychometric-feed at any time. Cronbach α per construct, IRT parameter summaries, pool distribution, virtues. Public read, no auth.

Call to Action

Direct. Try the API. Register a galaxy free.

Transitional. Read the IRT methodology (Phase 2). See the live pool lifecycle at /feed.

Spoke I/O (visual language v1)

Every toolbox spoke shares the same abstract choreography: typed inputs on the left, distilled verbs in the center, typed outputs on the right, and (when relevant) cross-spoke HTTP composition along the bottom rail. Source package: @people-analytics-toolbox/spoke-illustrations.

Try it now

Copy this curl. Paste in any terminal. Public read — no auth needed.

reincarnation.psychometric-feed

GET

Per-galaxy psychometric statistics. Public read. Returns Cronbach α, IRT a/b/c, virtues, pool counts.

curl -sS "https://people-analytics-toolbox.vercel.app/api/spokes/reincarnation/psychometric-feed"

Vendor the contract

The Zod contract is the source of truth. Vendor a copy into your consumer app — you keep it; we don't break it underneath you. Re-vendor when the version bumps.

// In your consumer app:
import { z } from "zod";

// Vendor a copy of these contracts from the toolbox repo at:
//   src/spokes/reincarnation/contracts/types.ts
// (consumers vendor, never import at runtime — keeps deploy boundary clean.)

import {
  AdaptiveSelectionRequestSchema,
  AdaptiveSelectionResponseSchema,
  PsychometricFeedResponseSchema,
  GalaxyRegisterRequestSchema,
  ResponseIngestRequestSchema,
  CONTRACT_VERSION,
} from "./vendored/reincarnation/types";

// Then call the toolbox over HTTP or MCP.
// See docs/EXTERNAL-CONSUMERS.md for onboarding.

Source path: src/spokes/reincarnation/contracts/types.ts · GitHub

Try in Dev Console Read the explainer See the live Insight Player Back to all spokes

Failure

Stale item banks silently poison trend lines; constructs look "stable" simply because discriminators flattened years ago—you cannot explain quarter-over-quarter divergence to leadership or auditors.

Success

Shorter, sharper surveys tied to latent information—not word count—plus live reliability and virtues your psychometric collaborators recognize. Explainable narratives when pulses shift because diagnostics are attributable to constructs and pools, not to dashboard noise.