{"id":"creative-thinking-for-research","name":"creative-thinking-for-research","summary":"創造的思考のための認知科学の枠組みをCSおよびAI研究のアイデアに応用します。組み合わせ的創造性、類推推論、制約操作、その他の実証的創造的戦略を活用し、真に新しい研究方向を求める際に活用します。","body":"# Creative Thinking for Research\n\nEight empirically grounded frameworks from cognitive science, applied to computer science and AI research. Unlike ad-hoc brainstorming, each framework here is backed by decades of creativity research — from Koestler's bisociation to Kauffman's adjacent possible. They target distinct cognitive operations: combining, reformulating, analogizing, constraining, inverting, abstracting, exploring boundaries, and holding contradictions.\n\n## When to Use This Skill\n\n- Generating genuinely novel ideas, not incremental extensions of prior work\n- Feeling trapped in a local optimum of thinking within a single subfield\n- Wanting to systematically apply creativity heuristics rather than waiting for inspiration\n- Preparing for a research retreat or PhD-level ideation session\n- Bridging between fields and seeking structural (not superficial) connections\n\n**Do NOT use this skill when**:\n- You need structured project-level brainstorming workflows (use `brainstorming-research-ideas`)\n- You have a well-defined problem and need execution help (use domain-specific skills)\n- You need a literature survey (use `scientific-skills:literature-review`)\n\n**Relationship to Brainstorm skill**: The brainstorm skill provides operational workflows (diverge → converge → refine) and practical filters. This skill provides the deeper cognitive engines that power creative leaps. Use them together: creative-thinking to generate raw insight, brainstorm to structure and evaluate it.\n\n---\n\n## Framework 1: Combinatorial Creativity (Bisociation)\n\nNovel ideas arise from combining existing concepts in unexpected ways. Arthur Koestler called this **bisociation** — connecting two previously unrelated frames of reference, as distinct from routine association within a single frame.\n\n**Why it works**: Meta-research consistently shows that breadth of knowledge is a precursor to creative output. People who read across disciplines produce more novel work. The combination itself is the creative act.\n\n**In CS Research**:\n- Biological evolution → optimization (genetic algorithms)\n- Game theory → networking (mechanism design for routing)\n- Statistical physics → machine learning (Boltzmann machines, energy-based models)\n- Linguistics → programming (type theory, formal grammars)\n\n**Systematic Bisociation Workflow**:\n\n1. **Select two domains** you have at least passing familiarity with\n2. **List core primitives** in each domain (5-10 fundamental concepts per domain)\n3. **Create a cross-product matrix**: row = concepts from Domain A, column = concepts from Domain B\n4. **For each cell**, ask: \"What would it mean to apply A's concept to B's problem?\"\n5. **Filter**: Which combinations produce a non-trivial, testable research question?\n6. **Validate structural depth**: Is the connection mechanistic or merely metaphorical?\n\n**Cross-Product Example**:\n\n| | Caching | Load Balancing | Fault Tolerance |\n|---|---------|---------------|-----------------|\n| **Natural Selection** | Evict least-fit entries | Adaptive allocation via fitness | Population-level redundancy |\n| **Immune Memory** | Learned threat signatures | Distributed detection | Self/non-self discrimination |\n| **Symbiosis** | Cooperative prefetching | Mutualistic resource sharing | Co-dependent resilience |\n\n**Quality Test**: A strong bisociation is not a surface metaphor (\"the network is like a brain\") but a structural mapping where the mechanism transfers (\"attention mechanisms implement a form of selective gating analogous to cognitive attention filtering\").\n\n**Self-Check**:\n- [ ] Is the connection structural (mechanisms map) or merely verbal (labels map)?\n- [ ] Does the combination generate testable predictions?\n- [ ] Would an expert in both fields find the connection non-obvious but sound?\n\n---\n\n## Framework 2: Problem Reformulation (Representational Change)\n\nGestalt psychologists identified that breakthroughs often come not from solving the problem as stated, but from **re-representing the problem itself**. Kaplan and Simon's work on insight shows that changing the problem space — the constraints, the abstraction level, the formalism — is often where creativity lives.\n\n**The Key Shift**: From \"How do I solve this problem?\" to \"Am I even thinking about this problem correctly?\"\n\n**Reformulation Strategies**:\n\n| Strategy | Example |\n|----------|---------|\n| **Change the objective** | \"Make the algorithm faster\" → \"Eliminate the need for this computation\" |\n| **Change the formalism** | Graph problem → linear algebra problem (spectral methods) |\n| **Change the granularity** | Per-token prediction → per-span prediction |\n| **Change the agent** | \"How should the model learn?\" → \"How should the data teach?\" (curriculum learning) |\n| **Change the timescale** | Real-time optimization → amortized inference |\n| **Invert the direction** | Forward simulation → inverse problem (learning from observations) |\n\n**Workflow**:\n\n1. State your current problem in one sentence\n2. Identify the **hidden assumptions** in that statement:\n   - What formalism are you using? (Could you use a different one?)\n   - What is the objective? (Is it the right objective?)\n   - What level of granularity? (Could you go coarser or finer?)\n   - Who is the agent? (Could you shift perspective?)\n3. For each assumption, **generate the alternative**: \"What if [opposite assumption]?\"\n4. For each alternative, ask: \"Does this reformulation make the problem easier, harder, or different in a useful way?\"\n5. A reformulation that makes a hard problem easy is often a publishable insight on its own\n\n**Classic CS Examples**:\n- **PageRank**: Reformulated \"find important web pages\" from content analysis to graph eigenvalue problem\n- **Dropout**: Reformulated \"prevent overfitting\" from regularization to approximate ensemble\n- **Attention**: Reformulated \"handle long sequences\" from remembering everything to selectively querying\n\n---\n\n## Framework 3: Analogical Reasoning (Structure-Mapping)\n\nDedre Gentner's **structure-mapping theory** and Kevin Dunbar's studies of real scientists show that analogy is the core engine of scientific creativity. The critical finding: surface-level analogies are common but weak; **structural or relational analogies** — where the deep causal/relational structure maps across domains — produce the most powerful insights.\n\n**Dunbar's Finding**: In the most successful labs, analogies from distant domains drove the most important discoveries. Nearby analogies refined ideas; distant analogies generated them.\n\n**Levels of Analogical Depth**:\n\n| Level | Description | Value | Example |\n|-------|-------------|-------|---------|\n| **Surface** | Things look similar | Low | \"A neural network is like a brain\" |\n| **Relational** | Relationships between entities match | Medium | \"Attention allocation in models parallels resource allocation in economics\" |\n| **Structural** | Deep causal mechanisms map | High | \"Diffusion models reverse a thermodynamic process; the math of non-equilibrium stat-mech directly applies\" |\n\n**Structure-Mapping Workflow**:\n\n1. **Describe your problem** using only relational/causal language (strip domain-specific nouns)\n   - Bad: \"We need to improve transformer attention efficiency\"\n   - Good: \"We have a system that must selectively aggregate information from a large set, where relevance is context-dependent and the cost scales quadratically with set size\"\n2. **Search for structural matches**: What other systems selectively aggregate from large sets?\n   - Database query optimization, visual attention in neuroscience, information retrieval, resource allocation\n3. **Pick the most distant match** with genuine structural fidelity\n4. **Map the solution mechanism**: How does the source domain solve this?\n5. **Transfer and adapt**: What changes when you bring that mechanism into your domain?\n6. **Generate predictions**: The analogy should tell you something you didn't already know\n\n**Validation Checklist**:\n- [ ] Does the mapping preserve causal/relational structure (not just labels)?\n- [ ] Can I identify at least one prediction the analogy makes in my domain?\n- [ ] Would an expert in the source domain confirm the mechanism is correctly understood?\n- [ ] Is the analogy non-obvious to my target audience?\n\n---\n\n## Framework 4: Constraint Manipulation (Boden's Framework)\n\nMargaret Boden's framework distinguishes three forms of creativity based on how they interact with constraints:\n\n| Type | Operation | CS Example |\n|------|-----------|------------|\n| **Exploratory** | Search within the existing conceptual space | Hyperparameter tuning, architecture search within a fixed paradigm |\n| **Combinational** | Combine elements from different spaces | Multi-task learning, neuro-symbolic methods |\n| **Transformational** | Change the rules of the space itself | Dropping the assumption that training requires labels (self-supervised learning) |\n\n**Transformational creativity is the rarest and highest-impact.** It happens when you change what is even considered a valid solution.\n\n**Constraint Analysis Workflow**:\n\n1. **List the constraints** of your current approach (5-10 constraints):\n   - Computational: \"Must fit in GPU memory\"\n   - Methodological: \"Requires labeled data\"\n   - Architectural: \"Uses fixed-length context\"\n   - Evaluative: \"Measured by accuracy on benchmark X\"\n2. **Classify each constraint**:\n   - **Hard**: Physically or logically necessary (cannot violate)\n   - **Soft**: Convention or historical accident (can question)\n   - **Hidden**: Not stated but implicitly assumed (most fertile for innovation)\n3. **For each soft/hidden constraint**, ask:\n   - What if we relaxed it? (streaming algorithms from relaxing \"fits in memory\")\n   - What if we tightened it? (efficiency research from tightening compute budgets)\n   - What if we replaced it with a different constraint entirely?\n4. **The most productive move** is often exposing and dropping a hidden constraint\n\n**Classic Examples of Constraint Transformation**:\n- \"Data must fit in memory\" → dropped → streaming algorithms, external memory\n- \"Training requires human labels\" → dropped → self-supervised learning\n- \"Models must be deterministic\" → dropped → variational methods, diffusion\n- \"Inference must happen in one pass\" → dropped → iterative refinement, chain-of-thought\n\n---\n\n## Framework 5: Negation and Inversion\n\nTake a core assumption in your field and negate it. This is formalized in De Bono's lateral thinking and the **TRIZ methodology** from engineering.\n\n**The Pattern**: \"What if [widely held assumption] is wrong, unnecessary, or invertible?\"\n\n**Systematic Negation Workflow**:\n\n1. **List 5-10 core assumptions** in your subfield (the things \"everyone knows\")\n2. **Negate each one** and ask: What system would you build?\n3. **Evaluate each negation**:\n   - Incoherent → discard\n   - Already explored → check if conditions have changed (see brainstorm skill, Framework 5)\n   - Unexplored and coherent → potential research direction\n\n**Negation Hall of Fame in CS**:\n\n| Assumption | Negation | Result |\n|-----------|----------|--------|\n| \"We need strong consistency\" | What if we don't? | Eventual consistency, CRDTs |\n| \"We need exact answers\" | What if approximate is fine? | Sketches, LSH, approximate nearest neighbors |\n| \"Labels are necessary\" | What if we learn without them? | Self-supervised learning, contrastive methods |\n| \"More parameters = more compute\" | What if we don't use all parameters? | Mixture of Experts, sparse models |\n| \"Training and inference are separate\" | What if the model keeps learning? | Online learning, test-time training |\n| \"Errors must be prevented\" | What if we embrace and correct them? | Speculative decoding, self-correction |\n\n**TRIZ-Inspired Principles for CS**:\n\n| TRIZ Principle | CS Application |\n|---------------|----------------|\n| **Inversion** | Reverse the process (generative vs. discriminative) |\n| **Segmentation** | Break monolithic into modular (microservices, mixture of experts) |\n| **Merging** | Combine separate steps (end-to-end learning) |\n| **Universality** | One component serves multiple functions (multi-task models) |\n| **Nesting** | Place one system inside another (meta-learning) |\n| **Dynamization** | Make static things adaptive (dynamic architectures, adaptive computation) |\n\n---\n\n## Framework 6: Abstraction and Generalization Laddering\n\nMoving up and down the abstraction ladder is a fundamental creative act. Polya's heuristics formalize this: *\"Can you solve a more general problem? A more specific one? An analogous one?\"*\n\n**Three Moves**:\n\n| Move | Question | Outcome |\n|------|----------|---------|\n| **Generalize** | \"Is my solution a special case of something broader?\" | Framework papers, unifying theories |\n| **Specialize** | \"What happens when I add extreme constraints?\" | Niche applications, surprising edge cases |\n| **Analogize** | \"Where else does this abstract pattern appear?\" | Cross-domain transfer (see Framework 3) |\n\n**Generalization Workflow**:\n1. State your specific result\n2. Replace each specific element with a variable: \"ResNet works for ImageNet\" → \"Architecture X works for distribution Y\"\n3. Ask: Under what conditions does this hold? What is the general principle?\n4. If the general principle is novel → that is the contribution\n\n**Specialization Workflow**:\n1. Take a general method\n2. Add extreme constraints: tiny data, huge dimensionality, adversarial inputs, real-time requirements\n3. Ask: Does the method still work? If not, why not?\n4. The failure case often reveals the method's true assumptions\n\n**When to Generalize vs. Specialize**:\n- Generalize when you have results but no explanation\n- Specialize when you have theory but no grounding\n- Analogize when you are stuck in either direction\n\n---\n\n## Framework 7: The Adjacent Possible (Kauffman / Johnson)\n\nStuart Kauffman's concept, popularized by Steven Johnson: innovation happens at the boundary of what is currently reachable — the **adjacent possible**. New ideas become thinkable once their prerequisites exist. This explains why simultaneous independent discovery is so common — multiple people reach the same boundary.\n\n**Practical Implication**: Map what has recently become possible and explore the space those enablers open.\n\n**Adjacent Possible Mapping Workflow**:\n\n1. **List recent enablers** (last 1-3 years):\n   - New hardware capabilities (longer context, faster inference, new accelerators)\n   - New datasets or benchmarks\n   - New open-source tools or frameworks\n   - New theoretical results\n   - New regulatory or social conditions\n2. **For each enabler, ask**: \"What was previously impossible or impractical that this now permits?\"\n3. **Combine enablers**: The most powerful adjacent possibles arise from the intersection of multiple new enablers\n4. **Check for competition**: If many people can see the same adjacent possible, speed or a unique angle matters\n\n**Current Adjacent Possibles (2025-2026)**:\n\n| Enabler | Newly Possible |\n|---------|---------------|\n| 1M+ token context windows | Full-codebase reasoning, book-length analysis |\n| Inference cost drops (100x in 2 years) | Real-time agentic loops, always-on AI assistants |\n| Open-weight models at GPT-4 level | Reproducible research on frontier capabilities |\n| Multimodal models (vision + language + audio) | Unified perception-reasoning systems |\n| Synthetic data at scale | Training data for domains with no natural data |\n| Tool-using models | Research automation, self-improving systems |\n\n**Timing Signal**: If your idea requires technology that doesn't exist yet, it's beyond the adjacent possible — park it. If your idea could have been done 5 years ago, someone probably did — check the literature. The sweet spot is ideas that became feasible in the last 6-18 months.\n\n---\n\n## Framework 8: Janusian and Dialectical Thinking\n\nAlbert Rothenberg's studies of eminent creators found that **holding two contradictory ideas simultaneously** is a hallmark of creative thinking. Named after Janus, the two-faced Roman god, this mode of thinking doesn't resolve contradictions by choosing a side — it generates new frameworks that transcend the opposition.\n\n**In CS**: The most influential results often emerge from tensions previously thought irreconcilable.\n\n| Contradiction | Resolution | Impact |\n|--------------|------------|--------|\n| Consistency AND Availability (distributed systems) | CAP theorem: formalized the trade-off, then Raft/CRDTs found practical middle grounds | Foundation of distributed systems theory |\n| Security AND Usability | Zero-knowledge proofs: prove knowledge without revealing it | Enabled private computation |\n| Expressiveness AND Tractability | Probabilistic programming: express complex models, automate inference | New programming paradigm |\n| Memorization AND Generalization | Grokking: models memorize first, then generalize with more training | New understanding of learning dynamics |\n| Compression AND Quality | Neural codecs that compress beyond information-theoretic limits via learned priors | Redefined compression research |\n\n**Dialectical Thinking Workflow**:\n\n1. **Identify a binary** in your field: A vs. B (two approaches, goals, or paradigms treated as opposites)\n2. **Resist choosing a side**. Instead ask:\n   - \"What would a system look like that achieves both A and B?\"\n   - \"Under what conditions is the A-B trade-off not fundamental?\"\n   - \"Is the opposition an artifact of how we formalized the problem?\"\n3. **Seek synthesis**: The resolution often requires a new abstraction that reframes the relationship\n4. **Test the synthesis**: Can you demonstrate empirically that both goals are achievable?\n\n**Self-Check**:\n- [ ] Am I holding the contradiction genuinely (not prematurely resolving it)?\n- [ ] Is the synthesis a new idea, not just a compromise (splitting the difference)?\n- [ ] Does the resolution change how people think about the problem, not just the solution?\n\n---\n\n## Combining Frameworks: A Creative Thinking Protocol\n\nThese frameworks are most powerful in combination. Here is a systematic protocol for a deep creative thinking session:\n\n### Phase 1: Map the Space (15 min)\n1. **Constraint Manipulation** (F4): List all constraints of the current paradigm. Mark which are hard, soft, hidden.\n2. **Adjacent Possible** (F7): List recent enablers that change the feasibility landscape.\n\n### Phase 2: Generate Disruptions (30 min)\n3. **Negation** (F5): Negate 3 soft/hidden constraints. What systems emerge?\n4. **Bisociation** (F1): Pick a distant field and create a cross-product matrix with your domain.\n5. **Problem Reformulation** (F2): Restate your problem 3 different ways (change objective, formalism, agent).\n\n### Phase 3: Deepen Promising Leads (30 min)\n6. **Analogical Reasoning** (F3): For each promising idea, find a structural analogy and extract predictions.\n7. **Abstraction Laddering** (F6): Move each idea up (generalize) and down (specialize).\n8. **Janusian Thinking** (F8): Identify any tensions. Can you synthesize rather than choose?\n\n### Phase 4: Evaluate (15 min)\nApply the two-sentence test (from the brainstorm skill):\n> \"**[Domain] currently struggles with [problem] because [reason].** We [approach] by [mechanism], which works because [insight].\"\n\nAny idea that survives all four phases and passes the two-sentence test is worth pursuing.\n\n---\n\n## Common Creative Blocks and Unblocking Strategies\n\n| Block | Symptom | Framework to Apply |\n|-------|---------|-------------------|\n| **Fixation** | Cannot stop thinking about the problem one way | Problem Reformulation (F2) — force a different representation |\n| **Tunnel vision** | All ideas come from the same subfield | Bisociation (F1) or Analogical Reasoning (F3) — import from elsewhere |\n| **Self-censoring** | Dismissing ideas as \"too weird\" before exploring | Negation (F5) — weird is the point; evaluate after generating |\n| **Incrementalism** | Every idea is \"+2% on benchmark X\" | Constraint Manipulation (F4) — change the rules, not the parameters |\n| **Analysis paralysis** | Too many options, cannot commit | Adjacent Possible (F7) — what is feasible right now? |\n| **False dichotomy** | Stuck choosing between two approaches | Janusian Thinking (F8) — seek synthesis, not selection |\n\n---\n\n## Usage Instructions for Agents\n\nWhen a researcher asks for help with creative thinking or novel ideation:\n\n1. **Assess the block**: What kind of thinking are they stuck in? (See Common Creative Blocks table)\n2. **Select 2-3 frameworks** based on the block type\n3. **Walk through each framework interactively**, asking the researcher to supply domain-specific content\n4. **Push for structural depth**: If an analogy or combination is surface-level, probe deeper\n5. **Maintain a running list** of all generated ideas, even unusual ones\n6. **Apply the two-sentence test** to candidates that survive exploration\n7. **Hand off to the brainstorm skill** for systematic evaluation (diverge → converge → refine)\n\n**Key Principles**:\n- Generative mode first, evaluative mode second — do not filter prematurely\n- Distant analogies are more valuable than nearby ones, but require more validation\n- The researcher's domain expertise is essential — the agent provides the cognitive scaffolding, not the domain knowledge\n- Encourage the researcher to sit with contradictions rather than resolve them quickly","author":"@OpenRaiser","ownerProfile":null,"authorContacts":null,"sourceUrl":"https://github.com/OpenRaiser/NanoResearch/tree/main/skills/vendor-ai-research/creative-thinking-for-research","license":"MIT","category":"review","lang":"en","tokens":4584,"stars":0,"calls30d":2,"claimed":false,"visibility":"public","origin":"crawler","version":"0.1.0","createdAt":"2026-08-22","updatedAt":"2026-08-22","files":[],"requires":{"mcp":[],"tools":[]},"safety":{"flags":[],"scannedAt":"2026-08-22","hasScripts":false,"networkEndpoints":[]}}