{"path":"research/meteor-prototype-analysis.md","content":"# Deliberus Meteor Prototype: Archaeological Database Analysis\n\n**Source:** delibr-meteor BSON dump (July 18, 2013, local archive)\n**Period of activity:** June 9 - July 16, 2013 (~5.5 weeks)\n**Platform:** Meteor.js, deployed at `deliberus2.meteor.com`\n\n---\n\n## 1. Executive Summary\n\nThe Meteor prototype was a **structured argumentation platform** with a surprisingly sophisticated data model for a 2013 prototype. It supported hierarchical argument trees, multi-axis voting (For/Against/Irrelevant), proposition substitution (rewording), evidence linking with source credibility voting, and user profiles. The database captures two genuine debates -- one on abolishing the Swedish monarchy, one on liberating North Korea -- conducted by 10 real users (plus 3 seed accounts) over 5.5 weeks.\n\n**Key findings:**\n- The **substitution/rewording mechanism** was the most novel and actively-used feature -- users refined each other's propositions to be more precise\n- Voting was **surprisingly balanced** (21 For vs 20 Against on arguments), suggesting genuine deliberation rather than echo-chamber dynamics\n- The system had **three-level depth**: propositions could have arguments, which could themselves have sub-arguments\n- Evidence linking to external sources (Wikipedia, DN.se) was implemented and used, with \"Found/Missing\" verification votes\n- Average element text length was **44 characters** -- short, punchy claims rather than paragraphs\n- The most active debate (\"Kungahuset bor avskaffas\") attracted **8 arguments** with **54 argument votes** from **6+ unique voters**\n\n---\n\n## 2. Database Schema Overview\n\n### 21 Collections (42 files total including metadata)\n\n| Collection | Documents | Purpose |\n|---|---|---|\n| **elements** | 33 | Core text atoms -- every claim, argument, and proposition is an element |\n| **propositions** | 32 | Wrapper around elements to track proposition-level voting & metrics |\n| **arguments** | 22 | Links connecting a support element to a top element (argument relationships) |\n| **argument_votes** | 54 | For/Against/Irrelevant votes on argument links |\n| **proposition_votes** | 25 | Yes/No/Abstain votes on propositions |\n| **substitute_propositions** | 8 | Proposed rewordings of propositions |\n| **substitute_proposition_votes** | 17 | Better/Worse votes on proposition rewordings |\n| **substitute_arguments** | 1 | Proposed rewordings of arguments |\n| **substitute_argument_votes** | 1 | Better/Worse votes on argument rewordings |\n| **evidences** | 3 | Links between elements and sources |\n| **evidence_votes** | 5 | Found/Missing verification votes on evidence |\n| **sources** | 3 | Source records with trustworthiness tracking |\n| **source_votes** | 5 | Trustworthy/Untrustworthy votes on sources |\n| **urls** | 3 | URL records for external references |\n| **references** | 1 | Anchor references within sources (e.g., Wikipedia section links) |\n| **profiles** | 4 | User profile metadata (only seed users) |\n| **users** | 13 | User accounts with Meteor SRP auth |\n| **argument_comments** | 1 | Comments on arguments |\n| **argument_comment_votes** | 3 | Like/Dislike votes on argument comments |\n| **system.users** | 2 | MongoDB system accounts |\n\n### Core Entity Relationships\n\n```\nElement (text atom)\n  |\n  +-- Proposition (wraps element, adds Yes/No/Abstain voting)\n  |     |\n  |     +-- SubstituteProposition (rewording: old element -> new element)\n  |\n  +-- Argument (topElementId -> supportElementId, adds For/Against/Irrelevant voting)\n  |     |\n  |     +-- SubstituteArgument (rewording)\n  |     +-- ArgumentComment (discussion on the argument itself)\n  |\n  +-- Evidence (links element to source + optional reference)\n        |\n        +-- Source (wraps URL, adds Trustworthy/Untrustworthy voting)\n              |\n              +-- Reference (anchor within source, e.g., Wikipedia section)\n```\n\n**Key architectural insight:** The `Element` is the fundamental atom. Everything else (propositions, arguments, evidence) references elements by ID. This means a single piece of text can simultaneously be a proposition AND an argument supporting another proposition. The data model is a **graph, not a tree**, even though the UI likely presented it as a tree.\n\n---\n\n## 3. Users\n\n### 13 total accounts: 3 seed + 10 registered\n\n**Seed accounts** (no `createdAt`, have `profile` field instead of `services`):\n- **Socrates** (`WEiEEogiaEw6yeF4m`) -- Most prolific seed creator (6 elements)\n- **Hegel** (`ecDw98vooNy8GsWJs`) -- 4 elements, started the monarchy debate\n- **Marx** (`jFfgxRMYm7i66fzfR`) -- 1 element (\"Staten kan ta over alla kungens slott\")\n\n**Registered accounts** (chronological):\n\n| Date | Username | ID | Notes |\n|------|----------|----|-------|\n| Jun 9, 15:27 | **Ptr** | CEYjF2Ydk7nhtnakh | Early adopter, colorful language |\n| Jun 9, 15:59 | **uzr** | up6rSwZGXbtf6Akms | Very active voter (12 votes), 5 elements |\n| Jun 11, 12:25 | **uzr2** | GFMfqb7bJgZoC3tEK | 2 votes only, likely a test account |\n| Jun 25, 10:23 | **nisse** | LDAGJRi3NxzSc6TdD | Most active voter (14 votes), Nils Janse? |\n| Jun 27, 10:44 | **fredrik** | j5TqBjPMJLb3SQZR4 | Fredrik Branstrom, co-founder |\n| Jul 2, 12:08 | **Tester** | mB74wctkzHwMfpw4P | 6 votes, likely a test/demo account |\n| Jul 2, 17:40 | **vsnellman** | iYtTzfq47DRvqGMiy | 1 element, 3 votes, provided academic-style argument |\n| Jul 4, 21:57 | **Mollgan** | mjWbRrvXHuv46Yp8N | Most prolific creator (7 elements), drove NK debate |\n| Jul 4, 21:59 | **Omniobjectmeter** | 7jgMyYrD8pvKXtJR6 | 3 elements, 7 votes, provocateur role (defended NK regime) |\n| Jul 6, 16:10 | **Pettsson** | doj5DXM2xto4cEEkB | Registered but no activity in dump |\n\n### Activity Distribution\n\n**Elements created per user:**\n| User | Elements | Notes |\n|------|----------|-------|\n| Mollgan | 7 | Most prolific content creator |\n| Socrates (seed) | 6 | Seed content + new contributions |\n| uzr | 5 | Active contributor |\n| Hegel (seed) | 4 | Seed content |\n| Omniobjectmeter | 3 | Provocateur role |\n| fredrik | 2 | Co-founder, moderate participation |\n| nisse | 2 | Focused more on voting |\n| Ptr | 2 | Quality over quantity |\n| vsnellman | 1 | Academic-style contribution |\n| Marx (seed) | 1 | Seed content only |\n\n**Votes cast (argument + proposition combined):**\n| User | Votes | Notes |\n|------|-------|-------|\n| nisse | 14 | Most engaged voter |\n| uzr | 12 | Heavy evaluator |\n| Mollgan | 9 | Balanced create + vote |\n| Ptr | 9 | Heavy evaluator |\n| Omniobjectmeter | 7 | Active participant |\n| Tester | 6 | Test account |\n| Socrates (seed) | 6 | Seed votes |\n| fredrik | 5 | Moderate engagement |\n| Hegel (seed) | 5 | Seed votes |\n| vsnellman | 3 | Minimal voting |\n| uzr2 | 2 | Minimal |\n| Marx (seed) | 1 | Seed only |\n\n**Pattern:** Creation and voting were done by different people. Mollgan created the most content but voted moderately. Nisse created only 2 elements but cast the most votes (14). This suggests the platform supported distinct **contributor** and **evaluator** roles naturally.\n\n---\n\n## 4. Debates (Reconstructed Argument Trees)\n\n### 4.1 Debate 1: \"Kungahuset bor avskaffas\" (The Swedish Monarchy Should Be Abolished)\n\n**The flagship debate.** Started by seed account Hegel. Attracted the most arguments (8), votes (54 argument votes), and participants. All content in Swedish.\n\n```\nROOT: \"Kungahuset bor avskaffas\" (The monarchy should be abolished)\n  Proposition votes: 4 Yes, 2 No\n  |\n  +-- FOR [net +7]: \"Det ger pengar till stadskassan\"\n  |     (It gives money to the city treasury)\n  |     7 For, 0 Against, 1 Irrelevant -- STRONGEST ARGUMENT\n  |     Evidence: Wikipedia \"Apanage\" article (1 Found vote)\n  |\n  +-- FOR [net +3]: \"Staten kan ta over alla kungens slott\"\n  |     (The state can take over all the king's castles)\n  |     3 For, 0 Against, 3 Irrelevant -- CONTROVERSIAL (high irrelevant)\n  |     |\n  |     +-- AGAINST [net -1]: \"Det strider mot aganderatten\"\n  |     |     (It violates property rights)\n  |     +-- AGAINST [net -1]: \"Det ar statens fastighetsverk som ager slotten\"\n  |           (The state property agency already owns the castles)\n  |\n  +-- FOR [net +2]: \"Kungahusets existens ar inte forenligt med principerna\n  |     pa vilka den liberala demokratin vilar\"\n  |     (The monarchy's existence is incompatible with liberal democratic principles)\n  |     2 For, 0 Against -- UNCHALLENGED PRINCIPLED ARGUMENT\n  |     By: vsnellman (the most academic-sounding contributor)\n  |\n  +-- FOR [net +1]: \"Kungen ar en dalig representant for Sverige\"\n  |     (The king is a poor representative for Sweden)\n  |     1 For, 0 Against, 2 Irrelevant\n  |     |\n  |     +-- FOR [net +1]: \"Det ar oserioso att tillsatta statschef genom knull\"\n  |           (It's unserious to appoint head of state through f***ing)\n  |           By: Ptr -- most colorful argument in the dataset\n  |\n  +-- AGAINST [net -4]: \"Kungen ar en bra representant for Sverige\"\n  |     (The king is a good representative for Sweden)\n  |     0 For, 4 Against, 2 Irrelevant -- STRONGLY REJECTED\n  |     This element had 4 SUBSTITUTE PROPOSITIONS (rewordings) -- most of any element\n  |     |\n  |     +-- AGAINST [net -3]: \"Kungen star inte for demokratiska varderingar\"\n  |           (The king doesn't stand for democratic values)\n  |           Evidence: Wikipedia Carl XVI Gustaf + Brunei state visit section\n  |           Evidence: DN article about scout medal to Saudi dictator\n  |\n  +-- AGAINST [net -2]: \"Det gynnar svensk export\"\n  |     (It benefits Swedish exports)\n  |     0 For, 2 Against, 1 Irrelevant\n  |     |\n  |     +-- FOR [net +1]: \"Kungen kan promota Sverige\"\n  |           (The king can promote Sweden)\n  |\n  +-- AGAINST [net -1]: \"Det ar fint med traditioner\"\n  |     (Traditions are nice)\n  |     0 For, 1 Against, 2 Irrelevant -- HIGH IRRELEVANT\n  |\n  +-- AGAINST [net -1]: \"Vi kan inte lamna Danmark som det enda landet\n        med en bra monarki\"\n        (We can't leave Denmark as the only country with a good monarchy)\n        1 Against, 0 For -- mildly sarcastic\n```\n\n**Observation:** The \"For\" side dominated with stronger arguments. The strongest argument (net +7, \"gives money to the treasury\") was backed by evidence. The strongest counterargument (\"good representative for Sweden\") was **rejected** by voters (net -4) and spawned 4 rewording attempts, suggesting users found the original formulation inadequate.\n\n### 4.2 Debate 2: \"Nordkoreas befolkning bor befrias\" (The North Korean People Should Be Liberated)\n\n**Started by Mollgan on July 4, 2013.** Shorter debate but demonstrates interesting dynamics. Omniobjectmeter played an explicit **devil's advocate** role, defending the North Korean regime.\n\n```\nROOT: \"Nordkoreas befolkning bor befrias fran den nu sittande regimens styre\"\n  (The NK people should be freed from the current regime's rule)\n  Proposition votes: 2 Yes, 0 No\n  |\n  +-- FOR [net +3]: \"Regimen bryter mot manskliga rattigheter\"\n  |     (The regime violates human rights)\n  |     3 For, 0 Against -- STRONGEST ARGUMENT\n  |     |\n  |     +-- AGAINST [net -1]: \"Inget klotter\" (No graffiti)\n  |     |     By: Mollgan -- tongue-in-cheek counterargument\n  |     +-- AGAINST [net -1]: \"Alla far komma till tals i landet\"\n  |           (Everyone gets to speak in the country)\n  |           By: Mollgan -- ironic/sarcastic\n  |\n  +-- AGAINST [net -1]: \"Manga manniskor skulle do vid en revolution\"\n  |     (Many people would die in a revolution)\n  |     0 For, 1 Against\n  |     |\n  |     +-- AGAINST [net -1]: \"Vill man gora en omelett far man knacka nagra agg\"\n  |           (If you want to make an omelet, you have to break some eggs)\n  |           By: Mollgan -- counter-counter argument\n  |\n  +-- AGAINST [net -1]: \"Juche stoder lokalproduktion vilket minskar\n  |     transportkostnader\"\n  |     (Juche supports local production which reduces transport costs)\n  |     By: Omniobjectmeter -- devil's advocate\n  |\n  +-- AGAINST [net -2]: \"Kim Jong Un ar en kompetent ledare\"\n        (Kim Jong Un is a competent leader)\n        By: Omniobjectmeter -- provocateur, rejected by voters\n```\n\n**Observation:** This debate demonstrates how the platform could handle **playful/ironic** argumentation alongside serious discourse. Mollgan's \"No graffiti\" and \"break some eggs\" arguments are clearly tongue-in-cheek. The system captured these without judgment, letting the voting mechanism express community assessment.\n\n---\n\n## 5. Substitute Propositions (Rewording Feature)\n\n**8 substitute propositions** were created, making this one of the most actively-used features relative to the dataset size. This is the most distinctive feature of the data model.\n\n### Rewording of \"Kungen ar en bra representant for Sverige\" (4 rewording attempts!)\n\n| New wording | By | Better | Worse | Net |\n|---|---|---|---|---|\n| \"Kungen ger en balanserad & rattvis bild av Sverige i media!\" | Socrates | 3 | 0 | **+3** (winner) |\n| \"I media gor kungen sig jattebra och promotar Sverige duktigt\" | Hegel | 2 | 1 | +1 |\n| \"Kungen ger en bra bild av Sverige i media\" | Hegel | 2 | 1 | +1 |\n| \"Kungahuset som funktion representerar Sverige bra\" | fredrik | 0 | 1 | -1 |\n\n**Pattern:** Users progressively refined a vague claim (\"the king is a good representative\") into increasingly precise formulations. The winning rewording added specificity (\"balanced & fair image in media\") while the losing one tried to abstract (\"the institution represents Sweden well\"). **The community preferred precision over abstraction.**\n\n### Rewording of \"Kungahuset bor avskaffas\" (2 attempts)\n\n| New wording | By | Better | Worse | Net |\n|---|---|---|---|---|\n| \"Kungen borde avga\" (The king should resign) | uzr | 1 | 1 | 0 |\n| \"Sverige borde ga over och bli en republik\" (Sweden should become a republic) | nisse | 0 | 1 | -1 |\n\n**Pattern:** Both rewordings tried to make the proposition less radical (abdication vs institution change), but neither gained community support. The original blunt formulation survived.\n\n### Rewording of \"Nordkorea bor befrias\" (2 attempts)\n\n| New wording | By | Better | Worse | Net |\n|---|---|---|---|---|\n| \"NK:s befolkning bor befrias fran den nu sittand regimens styre\" | Omniobjectmeter | 2 | 0 | +2 |\n| \"NK:s befolkning bor befrias fran den nu sittande regimens styre\" | Mollgan | 0 | 2 | -2 |\n\n**Pattern:** Omniobjectmeter added crucial precision -- \"the people should be freed from the regime\" vs \"NK should be freed\" (which could imply military invasion). The community recognized this as a meaningful semantic improvement. Mollgan's near-identical wording (fixing a typo) was rated worse, suggesting voters evaluated semantic value, not just correctness.\n\n### Substitute Arguments (1 instance)\n\nMarx proposed replacing \"Det ger pengar till stadskassan\" with \"Staten kan ta over alla kungens slott\" as a supporting argument. Voted \"Better\" by 1 user. This shows the substitution mechanism worked at both the proposition and argument level.\n\n---\n\n## 6. Evidence System\n\n**3 pieces of evidence** were linked, all to the monarchy debate.\n\n| Claim | Source | Anchor | Verification |\n|---|---|---|---|\n| \"Kungen uttalade sig positivt om diktaturen i Brunei\" | Wikipedia: Carl XVI Gustaf | #Statsbesoket_i_Brunei,_februari_2004 | 2 Found, 0 Missing |\n| \"Kungen gav en scoutmedalj till Saudiarabiens harskare\" | DN.se article | (none) | 2 Found, 0 Missing |\n| \"Det ger pengar till stadskassan\" | Wikipedia: Apanage | (none) | 1 Found, 0 Missing |\n\n**Sources** had trustworthiness voting:\n- Wikipedia Carl XVI Gustaf: 2 Trustworthy, 0 Untrustworthy\n- DN.se article: 2 Trustworthy, 0 Untrustworthy\n- Wikipedia Apanage: 1 Trustworthy, 0 Untrustworthy\n\n**Observation:** The evidence system was barely used (3 evidence entries out of 33 elements = 9% evidence coverage). But when it was used, the verification mechanism worked -- multiple users confirmed \"Found\" status. The `reference` field (Wikipedia section anchors) shows thoughtful granularity for a 2013 prototype.\n\n---\n\n## 7. Voting Patterns\n\n### Argument Votes (54 total)\n\n| Vote type | Count | % |\n|---|---|---|\n| For | 21 | 39% |\n| Against | 20 | 37% |\n| Irrelevant | 13 | 24% |\n\n**Remarkably balanced.** The near-equal For/Against distribution suggests genuine deliberation rather than dogpiling. The 24% \"Irrelevant\" rate is notable -- nearly 1 in 4 votes used this option, suggesting users valued the ability to flag arguments as off-topic without taking a side.\n\n### Proposition Votes (25 total)\n\n| Vote type | Count | % |\n|---|---|---|\n| Yes | 18 | 72% |\n| No | 6 | 24% |\n| Abstain | 1 | 4% |\n\n**Strong Yes bias.** This likely reflects that users voted on propositions they agreed with rather than systematically evaluating all propositions.\n\n### Substitute Proposition Votes (17 total)\n\n| Vote type | Count | % |\n|---|---|---|\n| Better | 10 | 59% |\n| Worse | 7 | 41% |\n\n**Surprisingly critical.** 41% \"Worse\" votes means the community actively rejected poor rewordings. This is a healthy signal -- the substitution mechanism wasn't just rubber-stamped.\n\n---\n\n## 8. Content Analysis\n\n### Element Text Characteristics\n\n- **Total elements:** 33\n- **Min length:** 13 characters (\"Inget klotter\")\n- **Max length:** 93 characters (\"Kungahusets existens ar inte forenligt med principerna pa vilka den liberala demokratin vilar\")\n- **Average length:** 44 characters\n- **Language:** 100% Swedish\n\n**Short, declarative claims.** The longest element is still under 100 characters -- about one tweet. This suggests the UI encouraged (or enforced?) brevity. The shortest elements tended to be tongue-in-cheek or lazy.\n\n### Topics\n\nTwo debates with clear thematic clusters:\n\n1. **Swedish monarchy** (most of the content): constitutional monarchy, democratic principles, representation, property rights, traditions, exports\n2. **North Korean regime** (secondary): human rights, revolution cost, Juche ideology, leadership competence\n\nBoth topics are **politically contentious but not partisan** in the Swedish context -- they cross traditional left/right lines. Good choices for testing a deliberation platform.\n\n---\n\n## 9. Data Model Insights\n\n### What Worked\n\n1. **Element-as-atom design**: Separating text content (elements) from structural roles (propositions, arguments) enabled text reuse across contexts. An element could be both a proposition AND a supporting argument for another proposition.\n\n2. **Multi-axis voting**: Different vote types for different contexts (Yes/No/Abstain for propositions, For/Against/Irrelevant for arguments, Better/Worse for substitutions, Trustworthy/Untrustworthy for sources, Found/Missing for evidence) showed nuanced engagement.\n\n3. **Substitution mechanism**: The most distinctive feature. Users genuinely engaged with it to refine formulations. The 4 rewording attempts for \"Kungen ar en bra representant\" show iterative precision-seeking.\n\n4. **Pre-computed indices**: `forIndex`, `againstIndex`, `controverseIndex`, `activityIndex`, `argumentedIndex` on propositions suggest real-time ranking/sorting was implemented. The `forIndex` formula appears to be `forVotes - againstVotes` (net score), and `controverseIndex` appears to penalize one-sided arguments.\n\n5. **Hierarchical arguments**: Arguments could themselves be argued (sub-arguments), creating depth. The monarchy debate shows 3-level depth in practice.\n\n### What Was Underused\n\n1. **Evidence system**: Only 3 evidence links across 33 elements. The friction of finding and linking sources was likely too high.\n\n2. **Argument comments**: Only 1 comment in the entire dataset. Discussion apparently happened through creating new arguments rather than commenting on existing ones.\n\n3. **Source credibility voting**: All 5 source votes were \"Trustworthy\" -- the \"Untrustworthy\" option was never used. Possibly because only Wikipedia and DN.se were cited.\n\n4. **Abstain voting**: Only 1 abstain vote out of 25 proposition votes. Users either agreed or disagreed; they rarely expressed uncertainty.\n\n### Schema Quirks\n\n- The `voter` field (early documents) was renamed to `userId` (later documents) -- a mid-development schema migration is visible in the data.\n- Some `votes` arrays contain `null` entries -- likely from deleted votes.\n- The `creator` field is sometimes a user ID, sometimes a username string -- inconsistent across seed and registered data.\n- `created` timestamps on seed data all share the same value (`1370760961223` = June 9, 02:56) -- bulk-imported.\n\n---\n\n## 10. Usage Patterns & Timeline\n\n### Activity Phases\n\n| Period | Activity |\n|---|---|\n| **Jun 9 (Day 1)** | Seed data loaded. Ptr and uzr register. First real votes. |\n| **Jun 9-11** | Ptr, uzr, uzr2 explore the monarchy debate. Ptr adds colorful arguments. |\n| **Jun 25** | nisse registers, becomes most active voter (14 votes over following weeks) |\n| **Jun 27** | fredrik registers, contributes rewordings |\n| **Jul 2** | Tester and vsnellman register. vsnellman adds the most academic argument. |\n| **Jul 4** | Mollgan and Omniobjectmeter register. North Korea debate starts. Most active day for new content. |\n| **Jul 6** | Pettsson registers (no activity captured) |\n| **Jul 16** | Last recorded activity (fredrik adds \"Vi kan inte lamna Danmark...\") |\n\n**Total active period: ~5.5 weeks** with activity concentrated in two bursts: initial launch (Jun 9-11) and the July 4 session that brought in new users and the second debate.\n\n### Engagement Decay\n\nThe database shows classic startup engagement decay:\n- **Week 1** (Jun 9-15): 4 registrations, heavy seeded content interaction\n- **Weeks 2-3** (Jun 16-30): 2 registrations, moderate voting activity\n- **Week 4** (Jul 1-7): 4 registrations, second content burst (NK debate)\n- **Week 5+** (Jul 8-16): 1 registration, trailing activity\n\n---\n\n## 11. Implications for Deliberus 2026\n\n### Features to Carry Forward\n\n1. **Short-form claims as atoms**: The 44-character average worked. Users naturally wrote concise, declarative statements. This aligns with the Claimify approach of atomic claim extraction.\n\n2. **Substitution/rewording mechanism**: The most novel and engaged-with feature. Users iteratively refined formulations toward precision. This is exactly the kind of collective knowledge refinement that Deliberus aims for.\n\n3. **Three-vote-type arguments** (For/Against/Irrelevant): The \"Irrelevant\" option was genuinely useful -- 24% of votes used it. This helps separate quality assessment from agreement/disagreement.\n\n4. **Evidence linking with verification**: The Found/Missing voting concept is sound. The low usage suggests the UX needs to be lower-friction (perhaps AI-assisted source finding?).\n\n5. **Pre-computed indices for ranking**: The `activityIndex`, `argumentedIndex`, `controverseIndex` show thinking about how to surface interesting propositions. The highest-activity proposition (\"Kungahuset bor avskaffas\") had `activityIndex: 6, argumentedIndex: 8, controverseIndex: -2`.\n\n### Lessons for the New System\n\n1. **The rewording data demonstrates emergent collective precision**: 4 people independently tried to make \"Kungen ar en bra representant\" more precise, and the community voted on which reformulation best captured the intent. This is **exactly the Deliberus thesis** -- structured argumentation produces better-articulated knowledge than freeform discussion.\n\n2. **Devil's advocate roles emerge naturally**: Omniobjectmeter's defense of the NK regime was clearly provocative but generated substantive counter-arguments. A good argumentation platform should welcome this.\n\n3. **Depth beyond 2 levels is rare**: Even in the most active debate, sub-sub-arguments were rare (only 4 instances of depth-3 argumentation). The UI/UX should optimize for 2-level depth with deeper nesting as an exception.\n\n4. **Evidence is the weakest link**: Only 9% of claims had evidence. AI-assisted evidence finding (Exa, Perplexity, etc.) could dramatically improve this ratio in 2026.\n\n5. **The voter/userId schema migration**: Even in a small prototype, schema evolution happened. The new system should plan for ontology evolution from day one.\n\n6. **Seed data was essential**: The three seed accounts (Socrates, Hegel, Marx) with pre-loaded content gave new users something to react to. An empty platform would have been far less engaging. Consider AI-seeded content for launch.\n\n---\n\n## Appendix A: All Elements (Complete Text)\n\n| ID | Creator | Text | Length |\n|---|---|---|---|\n| 9ZGHLHkR8SrtbyykB | Hegel | Kungahuset bor avskaffas | 24 |\n| zmM8ca2voqreCYnJX | Hegel | Kungen ar en bra representant for Sverige | 41 |\n| CYvWARkYQcduCqhWi | Hegel | Kungen ger en bra bild av Sverige i media | 41 |\n| PmfnKzEDDvqffbfxM | Hegel | I media gor kungen sig jattebra och promotar Sverige duktigt | 60 |\n| RsSwPSGdgut8Gj4mh | Socrates | Kungen star inte for demokratiska varderingar | 45 |\n| CgEvwtuSBtZJ3CuP2 | Socrates | Kungen uttalade sig positivt om diktaturen i Brunei | 51 |\n| ht7ENKrWHHtf3Krj4 | Socrates | Det ger pengar till stadskassan | 31 |\n| mFiSPnRmmEdQzbbhp | Socrates | Kungen gav en scoutmedalj till Saudiarabiens harskare, kung Abdullah | 68 |\n| WEuHXHx868qmXwfy3 | Socrates | Det ar inte sakert, kungahuset drivs till lag kostnad | 53 |\n| r8jbnsffqgCCDsNra | Socrates | Kungen ger en balanserad & rattvis bild av Sverige i media! | 59 |\n| h44BowmzCEqBkai58 | Marx | Staten kan ta over alla kungens slott | 37 |\n| K4LkmmvLrwZk7gxuT | Ptr | Det ar oserioso att tillsatta statschef genom knull | 51 |\n| j4EZ2SonBfrtEPEe7 | Ptr | Kungen ar en dalig representant for Sverige | 43 |\n| fTL2tvusyxvGdfbNo | uzr | Det ar fint med traditioner | 27 |\n| xTN9SEoos8ADxJXCJ | uzr | Det gynnar svensk export | 24 |\n| nG9abawHJ4qmDL36Y | uzr | Det strider mot aganderatten | 28 |\n| ZrvcwaZeZnMykyeWg | uzr | Det ar statens fastighetsverk som ager slotten | 46 |\n| py29Wp6GHCvBfFi7T | uzr | Kungen borde avga | 17 |\n| ZhwMLZwxb4xnYwBmK | nisse | Kungen kan promota Sverige | 26 |\n| EcAdMrDNs93aK3cuT | nisse | Sverige borde ga over och bli en republik | 41 |\n| pXompWCt52SDNgx6F | fredrik | Kungahuset som funktion representerar Sverige bra | 49 |\n| 6FSzfxZ8aZjTiJC2f | fredrik | Vi kan inte lamna Danmark som det enda landet med en bra monarki | 64 |\n| gTAX7biivNdCuKrGT | vsnellman | Kungahusets existens ar inte forenligt med principerna pa vilka den liberala demokratin vilar | 93 |\n| QwzaBeFK7nA4iKLvX | Mollgan | Nordkoreas befolkning bor befrias fran den nu sittande regimens styre | 69 |\n| H44xwJDBD4WY3PbdQ | Mollgan | Nordkorea bor befrias | 21 |\n| J5RdfY7QZWYTEgcMf | Mollgan | Regimen bryter mot manskliga rattigheter | 40 |\n| FqAkdYnGgExrDbW52 | Mollgan | Manga manniskor skulle do vid en revolution | 43 |\n| gvjaotmS7ogFeXbcC | Mollgan | Vill man gora en omelett far man knacka nagra agg | 49 |\n| TB6mhf6pihL9FK2pX | Mollgan | Inget klotter | 13 |\n| fg3rzb3f5qdzMrdTL | Mollgan | Alla far komma till tals i landet | 33 |\n| N6YCqWKoefgxvk4d6 | Omniobjectmeter | Kim Jong Un ar en kompetent ledare | 34 |\n| GMph9sX3abkAL6ECs | Omniobjectmeter | Juche stoder lokalproduktion vilket minskar transportkostnader | 62 |\n| PWQFSyR7NteDfX75x | Omniobjectmeter | Nordkoreas befolkning bor befrias fran den nu sittand regimens styre | 68 |\n\n## Appendix B: Source URLs\n\n1. `http://sv.wikipedia.org/wiki/Carl_XVI_Gustaf` (Trustworthy: 2/0)\n2. `http://www.dn.se/nyheter/sverige/kungen-slipper-kritik-for-medalj-till-diktator` (Trustworthy: 2/0)\n3. `http://sv.wikipedia.org/wiki/Apanage` (Trustworthy: 1/0)\n\n## Appendix C: Collection Schemas\n\n### elements\n```json\n{\n  \"_id\": \"string (Meteor ID)\",\n  \"text\": \"string\",\n  \"creator\": \"string (userId or username)\",\n  \"creatorName\": \"string (only seed data)\",\n  \"userId\": \"string (only registered users)\",\n  \"created\": \"number (Unix ms timestamp)\"\n}\n```\n\n### propositions\n```json\n{\n  \"_id\": \"string\",\n  \"elementId\": \"string (ref elements)\",\n  \"creator\": \"string\",\n  \"created\": \"number\",\n  \"votes\": [\"string (vote IDs)\"],\n  \"yesVotes\": \"number\",\n  \"noVotes\": \"number\",\n  \"abstainVotes\": \"number\",\n  \"activityIndex\": \"number\",\n  \"argumentedIndex\": \"number\",\n  \"controverseIndex\": \"number\",\n  \"substituteCount\": \"number\"\n}\n```\n\n### arguments\n```json\n{\n  \"_id\": \"string\",\n  \"topElementId\": \"string (the element being argued about)\",\n  \"supportElementId\": \"string (the element providing the argument)\",\n  \"propositionId\": \"string (ref propositions)\",\n  \"creator\": \"string\",\n  \"created\": \"number\",\n  \"votes\": [\"string\"],\n  \"forVotes\": \"number\",\n  \"againstVotes\": \"number\",\n  \"irrelevantVotes\": \"number\",\n  \"forIndex\": \"number (computed: forVotes - againstVotes - irrelevantVotes)\",\n  \"againstIndex\": \"number (computed: -(forVotes + irrelevantVotes))\",\n  \"commentCount\": \"number\",\n  \"substituteCount\": \"number\"\n}\n```\n\n### evidences\n```json\n{\n  \"_id\": \"string\",\n  \"elementId\": \"string (ref elements - the claim being evidenced)\",\n  \"sourceId\": \"string (ref sources)\",\n  \"referenceId\": \"string|null (ref references - anchor within source)\",\n  \"creator\": \"string\",\n  \"created\": \"number\",\n  \"votes\": [\"string\"],\n  \"foundVotes\": \"number\",\n  \"missingVotes\": \"number\",\n  \"foundIndex\": \"number\"\n}\n```\n\n### substitute_propositions\n```json\n{\n  \"_id\": \"string\",\n  \"oldElementId\": \"string (the element being reworded)\",\n  \"newElementId\": \"string (the proposed rewording)\",\n  \"creator\": \"string\",\n  \"created\": \"number\",\n  \"votes\": [\"string\"],\n  \"betterVotes\": \"number\",\n  \"worseVotes\": \"number\",\n  \"preferenceIndex\": \"number\",\n  \"relevanceIndex\": \"number\",\n  \"unrelatedVotes\": \"number\"\n}\n```\n\n---\n\n*Analysis generated March 27, 2026 from BSON dump dated July 18, 2013.*\n*21 collections, 196 total documents, ~41KB of raw BSON data.*\n"}