Relationship between DDS and the Deliberation Ontology (DEL)

[zhiganov]

One of this group's chairs has already published a validated ontology for deliberation data. With the specification being rewritten, it seems worth deciding explicitly whether DDS references it, maps to it, or stays separate — rather than arriving at an answer by default.

What exists

DEL, the Deliberation Ontology — @Stocastico96 (CIRSFID-ALMA AI, Bologna) with Víctor Rodríguez-Doncel (OEG, Universidad Politécnica de Madrid).

• OWL 2, persistent URI https://w3id.org/deliberation
• Built with the LOT methodology, reusing several existing ontologies
• Validated with the OOPS! scanner
• Three models: Process (how deliberation unfolds), Participant (who takes part), Argument (what gets said, including support and attack relations)
• Positioned against prior work including DELIB, AIF, AMO, SIOC, LKIF and PAKT

Per the version presented at Citizen Infrastructure Builders Collective's Demo Day on 2026-07-03, it has been applied to parliamentary transcripts, civic platform proposals and forum discussions across several platforms. Simone can speak to the current state better than I can.

The question

If DDS is moving toward protocol- and schema-agnostic guidance, then "what shape is deliberation data" is a question it will have to answer or defer somewhere. Three options, and the group may well have already settled on one informally:

1. Reference DEL as the recommended data model, with DDS covering transport, provenance and conformance over it.
2. Map to it — DDS keeps its own vocabulary but publishes a correspondence, so data can round-trip.
3. Stay separate, on the grounds that DDS's scope genuinely differs — reasonable, but better recorded than assumed.

The cost of picking is low right now and rises once either side has implementers.

Why raise it

Reusing a published, validated ontology with a persistent URI is cheaper than writing one, and DEL's argument model in particular covers ground a deliberation stack would otherwise have to invent. It also reads oddly from outside for a group to write a data standard while one of its own chairs maintains a published ontology for the same domain, with no stated relationship between them.

Happy to be told this is already resolved somewhere I haven't looked.

[Stocastico96]

Thanks for raising this explicitly — you're right that it's better decided than defaulted. Disclosure first: I'm both a chair of this group and the lead author of DEL, so what follows is a proposal for the group to decide formally, ideally at the next monthly call.

The honest answer to "does the ontology reflect the actual group view?" is: not yet — because the group has never been asked. Rather than retrofitting legitimacy onto DEL, I propose we fix the underlying gap: the group has no explicit mandate about its outputs. So:

Proposal: the group adopts a mandate to produce two (possibly three) specifications, each with its own editors, following the model of successful Community Groups like DPVCG (five multi-authored reports) and OntoLex:

Deliberation Data Model — an ontology/vocabulary specification. I propose DEL as the starting text (rationale below), with an explicit reconciliation task: integrating the best of the existing initiatives, currently at least three — DEL, Metagov's Deliberation Interoperability Project (see Stormonth-Darling & Barry, Towards Interoperability, Aarhus 2025, DOI 10.1145/3737609.3747119), and the EU AI4Deliberation project's knowledge-graph work (IBIS core, xAIF argument interchange). Their contributors would be invited as co-editors — ontologies are consensual artifacts, and a spec with many parties behind it is a stronger spec.
Protocols & Software — transport, identity, anonymity (the current spec documents), plus the practical interop tooling: Metagov's Rust adaptors for Polis/HeyForm/Talk to the City are genuinely valuable and could be adapted to read/write the shared data model rather than a parallel one.
(Optional) Use Cases & Requirements — if the group feels the first two need grounding.
Why DEL as the starting text for (1): it's the only one of the three that is a published, validated OWL ontology with a persistent dereferenceable namespace — v1.1.0 released this month at https://w3id.org/deliberation/ontology (LOT methodology, OOPS!-validated, CC BY 4.0, public SPARQL endpoint, browsable documentation), applied to EP debates, EU Have Your Say, Decide Madrid, Decidim, DeliData, Pol.is, CIP Global Dialogues and US Supreme Court arguments. 1.1.0 added ArtificialAgent, Facilitator and Reaction precisely as a result of comparing against Metagov's design, and SKOS alignments to their exported terms are published. There's also a self-service mapping tool for any platform's data model (exports SSSOM/YARRRML). That said: starting text, not final word — if reconciliation changes it, that's the process working.

On IPR, to pre-empt the obvious question: adopting a data-model spec as a CG deliverable transfers no copyright. DEL is already CC BY 4.0 and remains © its authors (University of Bologna / UPM); the CG CLA layers royalty-free licensing and patent commitments on future contributions — which is what implementers need. What changes is change control: the vocabulary evolves by group consensus, with named editors. I believe that's the right structural answer to the "reads oddly from outside" point — which, I agree, it currently does.

One concrete coordination item that can't wait for the reconciliation to finish: DEL and DIP are both minting AT Protocol lexicons in the same org.deliberation.* NSID namespace with different record shapes for overlapping concepts, and given our ATProto PoC this will surface in DDS implementations first. I'm contacting the Metagov folks to coordinate namespace stewardship and will report back here.
@stuartlynn @salteelee

Concretely: (1) mandate + deliverables list on the agenda of the next call, outcome recorded in spec/design-rationale.md whatever it is; (2) if adopted, we formally invite Metagov and AI4Deliberation contributors as co-editors of the data-model spec. Happy to draft both documents; the decisions should be the group's, not the chair's.

[nicobao]

Hi, thanks for bringing that up! Very interesting discussions.

On my side I am nothing but an expert in Semantic data/Ontology. I just know why it's useful and we should probably get them right.

I think, this is a very complex discussion, that should be done within the DDS Subgroup related to "Semantic data/Ontology".

And then, anyone interested would join. Either as an expert, like Simone, or as a technical "user", like me.

Discussions would be open and inform to the larger group.

[nicobao]

Anyone interested is invited to join https://luma.com/bhta7dic to discuss the matter