creative-thinking-for-research_skill

Overview

This skill applies eight empirically grounded cognitive-science frameworks to spark genuinely novel computer science and AI research directions. It packages structured heuristics—combinatorial combination, analogical mapping, constraint manipulation, inversion, abstraction laddering, and more—so you can move beyond incremental tweaks. Use it to generate research hypotheses that are mechanistic, testable, and non-obvious.

How this skill works

The skill inspects your current problem statement and cognitive framing, then systematically applies different creative engines: bisociation (cross-domain combination), representational change, structural analogy, constraint analysis, negation/inversion, abstraction/generalization, and exploration of the adjacent possible. Each framework provides a short workflow: how to transform assumptions, produce candidate hypotheses, and validate structural depth and testability. Outputs are filtered by novelty, mechanistic plausibility, and feasibility for follow-up experiments.

When to use it

• When you want genuinely novel research ideas rather than incremental improvements
• If you feel stuck in a local optimum inside a single subfield
• Preparing for a deep ideation session, grant pitch, or PhD research outline
• When bridging disciplines and seeking structural (not superficial) connections
• To expose hidden assumptions that, when changed, yield high-impact directions

Best practices

• Start with a one-sentence problem statement and list hidden assumptions before applying frameworks
• Combine frameworks: use bisociation to generate candidates and representational change to reformulate them
• Prefer structural mappings over surface metaphors; require at least one testable prediction
• Classify constraints as hard/soft/hidden and target hidden ones for transformational moves
• Iterate rapidly: generate many low-cost hypotheses, then filter for mechanistic depth and feasibility

Example use cases

• Cross-domain idea generation: map immunology primitives to distributed systems to design adaptive fault tolerance
• Problem reformulation: turn a compute-bound objective into one minimizing required computation by changing formalism
• Analogical transfer: use resource allocation models from economics to redesign attention for sparse retrieval
• Constraint-dropping: challenge the assumption of labeled data to derive self-supervised protocols for a new modality
• Inversion exploration: negate 'training and inference are separate' to propose online, continual learning pipelines