Research / analysis

Current research

CA-R3 — selection technology deep assessment

Type: analysis Updated: 2026-10-05 Scale: cross-jurisdictional Path: technology/SELECTION-TECHNOLOGY.md

technologyselectionsortition

Sources: SRC-0073SRC-0074SRC-0075SRC-0076SRC-0077SRC-0078SRC-0083SRC-0144SRC-0145SRC-0156SRC-0152

Selection technology deep assessment

CA-R3 record. Updated: 2026-10-05. Compares OpenDLP, sortition-algorithms, Panelot against CA-R2 selection requirements.

Layer separation (important)

Four distinct layers are frequently conflated:

  1. Selection algorithm — the fairness/optimisation mathematics (leximin etc.).
  2. Civic-lottery workflow — respondent pool, targets, run, replacements, member handover.
  3. Governance specification — the authorised artefact saying which strata/quotas/weights (CA-REQ-010/011).
  4. Audit ceremony — seed capture, witness record, reproducibility bundle, publication (CA-REQ-013/014).

The corpus shows algorithms 1 is solved and published; 2 exists in one early platform; 3 and 4 are under-served everywhere and are mostly institutional-method shaped.

Comparison

AspectOpenDLPsortition-algorithmsPanelot
Layerworkflow (2) + engine via depengine (1)workflow+engine, hosted
Stratified modeltargets with min/max categoriescategory quotas inputleximin per published method
Optimisation objectivevia lib: maximin/leximin/nash/diversimaxthe family itselfleximin (Nature paper lineage)
Randomness/seednot documented publiclycaller-supplied seed semantics (Python random)opaque
Reproducible runsrun records persisted (targets_used, remaining_ids)deterministic given inputs+seedclaimed by method; unverifiable hosted
Witness/public lotterynot evidencedn/a (library)not evidenced
Audit artefactsDB records; report CSV in dev historynone emittedunknown
Reserves/replacementfirst-class operationremaining-pool supporthosted feature, undocumented
Privacy modelrespondent CSVs to operator; engine sees anonymised featuresanonymised feature matrixinside hosted service
Input/outputCSV in; panel + report outPython API / tabularCSV in/out (per practitioner material)
Algorithm transparencytransitively full (open lib)full source + papermethod published, implementation not
LicenceApache-2.0GPL-3.0unasserted (no public source)
Self-hostyes (Docker, docs)library — triviallyno
Ops complexitymoderate (Flask+Celery+Redis+PG)none (in-process)none (SaaS)
Pinned refHEAD 4d73880 (2026-10-05), no tagsv0.12.14 (2026-10-01)n/a

Findings

Answer: does Cicada need its own selection algorithm?

No — evidence does not justify it. CA-R2 requirements are satisfiable by the published engine. What no existing system supplies is the governance-specification artefact and the witness/reproducibility bundle — thin layers Cicada likely must own regardless of engine choice. Engine selection between adapt (wrap sortition-algorithms) and integrate (call OpenDLP’s workflow) is a CA-A1 decision, not needed now.

Sources