Software — OpenDLP (Open Democratic Lottery Platform)
Classification: civic-lottery platform (invitation→registration→selection→replacement workflow) around a stratified-selection engine. Updated: 2026-10-05. Evidence labels per RESEARCH-METHOD.md. Fit (CA-R1 boundary): unresolved — candidate for later CA-R3 assessment. No reuse/adapt/integrate decision made.
Identity
- Project: OpenDLP — Open Democratic Lottery Platform.
- Organisation/maintainer: Sortition Foundation (see
research/organisations/sortition-foundation.md). - Repository: https://github.com/sortitionfoundation/opendlp — Python; created 2025-07-16; actively pushed as of 2026-10-04; small community footprint (2 stars, 0 forks, 7 open issues) — early maturity (SRC-0073).
- Hosted instance + docs: democraticlottery.org (login) and docs.democraticlottery.org (SRC-0074).
- Repo structure evidence:
backend/,docs/,.adr-dir/(architecture decision records), pre-commit/secret-scanning config (SRC-0073).
Licence and distribution
- Licence: Apache-2.0 (repo metadata; SRC-0073). Genuinely open source.
- Hosted SaaS operated by Sortition Foundation; self-hosting appears possible from the public repo but packaged deployment documentation was not verified in this pass (Open question).
Core purpose
- Claimed: “streamline democratic lotteries” for citizens’ assemblies and mini-publics (SRC-0070).
- Observed: documentation covers the operator workflow — invitation to register interest, registration page, running selections, running replacements, keeping selected people warm/confirmed (SRC-0074, SRC-0075). Assembly creation, respondent CSV format and target CSV format are documented (SRC-0075).
Selection / sortition model
- Observed: stratified random selection of a fixed-size panel from an opt-in respondent pool against operator-defined targets (categories with min/max), using the Sortition Foundation
sortition-algorithmslibrary — the algorithm family developed with Harvard/Carnegie Mellon academics and published in Nature (SRC-0076, SRC-0077, SRC-0078). The library implements maximin, leximin, nash, diversimax and legacy methods, with reporting of quota fulfilment (SRC-0077). - Observed: replacements are a first-class documented operation (reserve pool → replacement run) (SRC-0074).
- Invitation distribution (postal letters etc.) is described as an operational stage in the docs, i.e. largely outside the software — the platform manages respondents and selection (SRC-0074, SRC-0075). Analysis: confirms the invitation/mailing layer is institution+ops work, not code.
Audit / reproducibility
- Observed: the underlying selection code is open source with a documented walkthrough enabling operators to run a selection independently (SRC-0076, SRC-0077). Random-seed handling, per-run audit artefacts and witness/attestation mechanics are not documented in the reviewed material (Open question — compare Decidim’s explicit die-roll+witnesses seed record, SRC-0083).
Identity and privacy
- Observed: a “Data and the law” docs section exists but is marked “still to be written”; a User Data Agreement page exists (SRC-0074, SRC-0099). Data-handling assumptions therefore partially documented; GDPR material not yet published (Open question).
API / export / interoperability
- Observed: CSV-based respondent and target inputs documented (SRC-0075). No public API documentation found in reviewed material (Open question). Docs repository is public: github.com/sortitionfoundation/opendlp-docs (SRC-0074).
Accessibility / participation constraints
- Analysis: the platform is an operator tool; participant-facing accessibility sits in the invitation/registration layer, which the docs treat as part postal/offline (letters, phone registration in Sortition Foundation practice, SRC-0038, SRC-0070).
Deployment evidence
- Observed/Claimed: operated by Sortition Foundation, which reports >200 democratic lotteries; OpenDLP itself is the newer platform (repo created 2025-07) — extent of production use vs the older
stratification-appworkflow is unclear (SRC-0070, SRC-0073, SRC-0040).
Limitations / open questions
- Early-stage platform; data-protection documentation incomplete (SRC-0099).
- Seed/witness/audit-artefact model not evidenced — central to Cicada’s reproducibility requirements.
- Boundary between OpenDLP and the separate
sortition-algorithms/stratification-appcodebases needs CA-R3 verification.
Sources
- SRC-0073 (repo metadata), SRC-0074 (docs root), SRC-0075 (running selections), SRC-0076 (selection algorithms page), SRC-0077 (sortition-algorithms repo), SRC-0078 (Nature paper), SRC-0099 (data-and-legal status), SRC-0040, SRC-0070 (institutional context).
CA-R3 pinned assessment
- Pinned source/ref:
sortitionfoundation/opendlp@4d73880c6423716f68c6de0c9d0be777b71974a1(SRC-0144); docsopendlp-docs@02191d2f57059fd37a973b0d845ea0104ef75e4b(SRC-0151). - Licence: Apache-2.0.
- Maintenance: active young project; deployment documentation is still explicitly incomplete in the repository README.
- Deployment model: Python web application; hosted workflow exists; self-host is plausible from source but not mature enough to assume as a turnkey operational package.
- Relevant CA-REQ coverage: strong/partial for CA-REQ-005–008, 010–019 and 025; weak for downstream agenda/deliberation/response.
- API/integration seam: CSV respondent/target workflow is evidenced; stable public API documentation was not found.
- Audit/provenance: algorithm is open, but CA-REQ-013/014 seed+witness+immutable-run artefacts are not sufficiently evidenced in OpenDLP itself.
- Identity/privacy: stores respondent data; data-protection documentation remains incomplete.
- Operational burden: moderate; narrower than a general participation platform.
- Extension/customisation: Apache-2.0 permits modification, but a long-lived fork would add maintenance burden.
- Accessibility/offline: offline invitation/registration can be supported institutionally; software does not itself satisfy non-digital parity.
- Material gaps/adoption risks: audit ceremony, API contract, mature self-host operations.
- Provisional classification: integrate.