{"path":"sketches.md","content":"# Deliberus Sketches & UI Concepts\n\nThis document catalogs the hand-drawn sketches and UI concepts shared across conversations, serving as the visual design record for Deliberus.\n\n## Sketch 1: Contention Diagram\n\n**Source**: Hand-drawn on white paper (uploaded to ChatGPT and Claude, Mar 2026)\n\n**Description**: Central \"Contention\" node with arrows flowing in from multiple directions:\n- **Upper area**: Cluster of argument cards grouped in a bubble enclosure, with a \"Counter argument\" node explicitly labeled\n- **Lower area**: \"Premise\" node surrounded by its own cluster of supporting cards\n- **Left and right**: Additional argument clusters in bubble enclosures, each containing 3-5 rectangular nodes\n- **Arrows**: Flow inward toward the Contention from all surrounding clusters\n- **Bottom right**: Three scoring axes marked as triangles — **truthiness**, **relevance**, **controversiality**\n\n**Design implications**:\n- Arguments are spatially clustered, not flat lists\n- Counter-arguments are visually distinct from supporting arguments\n- Multiple dimensions of evaluation are visible simultaneously\n- The \"bubble\" grouping suggests argument bundles are first-class UI objects\n\n## Sketch 2: Argument Bundles / Conclusion\n\n**Source**: Hand-drawn on white paper (uploaded to ChatGPT and Claude, Mar 2026)\n\n**Description**: Top to bottom layout:\n1. **\"Conclusion\"** node at top (rounded rectangle)\n2. Three **\"VALID\" badges** connecting conclusion to argument bundles below\n3. Three named argument bundles:\n   - \"speed argument\" (left) — 3-4 premise cards\n   - \"dream argument\" (center) — 3-4 premise cards\n   - \"cobalt argument\" (right) — 3-4 premise cards with a small triangle marker\n4. **Tab navigation bar** at bottom: `speed | dream | cobalt | sweet` with cursor clicking \"dream\"\n\n**Design implications**:\n- Multiple independent lines of reasoning can support the same conclusion\n- \"VALID\" badges suggest entailment checking (Tier C argument scheme validation)\n- Tab navigation allows switching between bundles without losing the overview\n- This is a KEY UX innovation — bundles as semi-contained threads that resolve into the same conclusion\n\n## Sketch 3: Multi-axis Scoring / 30°C Example\n\n**Source**: Referenced in ChatGPT brief from notebook sketches\n\n**Description**: Rating axes shown as triangles/sliders:\n- TRUE / FALSE\n- \"follows / does not follow\" (logical entailment)\n- Example: \"It's 30°C today\" → \"It's hot outside?\" (claim vs inference)\n- Vote counts: \"250 total\" with percentages\n- Categories: \"Most negative\", \"Most supportive\"\n\n**Design implications**:\n- Distinguishes measurable statements from value-laden classifications\n- Full vote distributions visible (not just averages)\n- Surfaces strongest pro/con, not just majority opinion\n\n## Sketch 4: Feed Algorithm Comparison\n\n**Source**: Referenced in ChatGPT brief from notebook page\n\n**Description**: Contrasts two approaches:\n- **Infinite scroll** (left side)\n- **\"organic (redraw)\"** (right side) — with funnel shape showing \"60%\" threshold\n\n**Design implications**:\n- Content should be filtered/weighted by relevance/value density\n- System continuously recomputes salience as new info arrives\n- Not a dump of the whole graph — intelligent surfacing\n\n## Sketch 5: Felicific Calculus / Values\n\n**Source**: Referenced in ChatGPT brief from notebook\n\n**Description**:\n- \"felicific calculus?\"\n- \"values for every proposition?\"\n- \"should / ought\"\n- \"How to determine — equations — more complicated than that\"\n\n**Design implications**:\n- Normative claims get separate treatment\n- Option for computation, but wariness of naive math over ethics\n- Values may be quantified, qualified (contextual), or aggregated — but carefully\n\n## Sketch 6: Similarity / Duplicate Detection\n\n**Source**: Referenced in ChatGPT brief from notebook\n\n**Description**:\n- Similar existing points should be visually available and interconnected\n- Accessible instantly when someone begins writing similar words\n- Goal: reduce repetition and rehash\n\n**Design implications**:\n- Composing is also retrieval + linking, not freeform posting into void\n- As user types, system suggests: \"this claim already exists\", \"here are its strongest arguments/counters\", \"here's the current belief distribution\"\n\n## Sketch Grid (from ChatGPT analysis)\n\nThe ChatGPT conversation identified ~13 notebook/sketch photos with these themes:\n- Tree-based epistemic visualizations\n- Semantic clustering of arguments\n- Truthiness vs controversy weighting axes\n- Zoom mechanics, context definitions, and node voting\n- Visual grammars for rationality and validity\n- Emotion-tagged or affect-layered representations\n- Collective cognition primitives (asynchronous suggestion routing)\n\n**Assessment**: \"Some of these remind me of early concept art for tools like Polis, Kialo, and debate mapping systems — but more multidimensional.\"\n\n## Sketch 7: Lean vs Strategy + Concept Tracking (2013)\n\n**Source**: Photo from FERMI drive, uploaded to Claude Mar 28, 2026. Large handwritten sheet on white paper.\n\n**Left section — \"LEAN vs STRATEGI\":**\n- **Lean route**: Steg 1 (Step 1) → PDF ingestion → extract claims → Steg 2 (Step 2). \"lukermodell\" (lurker model) for users. Percentages (50%, 20%) for participation tiers\n- **Strategic route**: richer semantic modeling, value/ethics computation, social-force detection, institutional engagement\n- **\"axiom 100% (grundläggande)\"**: Axioms labeled as fundamental with 100% confidence — the value bedrock\n- **\"Politiskt bevis/data\"** (Political evidence/data): evidence sources with scoring\n- **\"Crosspollinering\"** (Cross-pollination): interactions between user groups\n\n**Right section — \"Concept tracking (experimental)\":**\n- **Word disambiguation example**: \"Sweet\" → \"Sweet (sugar)\" / \"Sweet (nice)\" / \"Sweet (?)\" — a single word decomposed into distinct sense-nodes\n- **Concept branching**: \"humans\" tracked with branching meanings (\"humans kan have feelings\" 70%, \"humans have skin\" 80%)\n- **Claim examples**: \"metacharaktherise dinosaur Fred\" — applying properties to entities\n- **Crowdsourced operations**: \"Meta: merges/splits\" — community can merge equivalent concepts or split ambiguous ones\n- **\"Rösta/argumentera\"** (Vote/argue): voting and argumentation as disambiguation mechanisms\n- **Lifecycle states**: \"Stabilisera\" (Stabilize), \"Sidouppload\" (Side-upload), \"Komplett\" (Complete)\n- **Renaming operations**: \"rename - reverse/rename(?)\" for concept evolution\n\n**Design implications**:\n- Concept tracking (semantic disambiguation) was conceived as a first-class experimental feature in 2013\n- Directly influenced by Emanuel Kumlien's linguistic insights about discourse layers and meaning ambiguity\n- The \"Sweet\" example demonstrates treating word senses as SEPARATE nodes — exactly the definitional-claim-as-first-class-node design derived independently in Session 3 (2026)\n- Merge/split operations make disambiguation a CROWDSOURCED activity, not a top-down classification\n- Axioms are explicitly distinguished as \"100% fundamental\" — the irreducible bedrock of worldviews\n\n**See also**: [research/semantic-disambiguation-and-concept-tracking.md](research/semantic-disambiguation-and-concept-tracking.md) · [research/voice-memo-emanuel-sofia.md](research/voice-memo-emanuel-sofia.md)\n\n---\n\n**Note**: Original sketch images are in ChatGPT conversation history and uploaded to Claude (Mar 27, 2026). Consider saving high-resolution versions to `docs/sketches/` for permanent reference.\n\n**See also**: [Object Model](object-model.md) · [Vision](vision.md) · [Technical Direction](technical-direction.md)\n"}