From Theory to Protocol: Executable Frameworks for Creative Emergence and Strategic Foresight

Shun Fujiyoshi

cs.HCMar 6, 2026

From Theory to Protocol: Executable Frameworks for Creative Emergence and Strategic Foresight

Authors: Shun Fujiyoshi

Organizations: Independent Researcher

Abstract

Creativity and strategic foresight have been extensively studied through descriptive theories -- Koestler's bisociation (1964), de Bono's lateral thinking (1967), and Ansoff's weak signals (1975) explain why creative and strategic insights occur, but offer limited guidance on how to produce them on demand. This paper presents two executable protocols that bridge this theory-practice gap: GHOSTY COLLIDER, a 5-step protocol for cross-domain creative emergence through structural de-labeling and collision, and PRECOG PROTOCOL, a 5-step protocol for signal-based strategic foresight with multi-axis timing judgment. We formalize established theories into repeatable, step-by-step procedures with explicit quality criteria, anti-pattern detection, and measurable outputs. We evaluate the protocols through three complementary methods: (1) five detailed case studies across distinct domains, (2) controlled comparisons against standard methods using identical inputs, and (3) a batch experiment across eight random domain pairings (N=8, success rate 87.5%, failure rate 12.5%) with one blind evaluation. Preliminary evidence suggests that protocol-driven outputs exhibit greater structural novelty, higher parameter specificity, and qualitatively distinct creative directions compared to outputs from standard methods. The blind evaluation confirmed the direction of author assessments (protocol output scored 74/80 vs. brainstorming 49/80). These results, while limited by single-operator execution, indicate that the theory-to-protocol translation preserves and potentially enhances the generative power of the underlying theories. The protocols, updated to version 2 incorporating lessons from failure case analysis, are released as open-access documents under CC BY-NC 4.0 at https://github.com/GhostyAI-HA/ghosty-collider.

Figures & tables

Anti-Pattern	Description	Remedy
Shallow Ghosting	Ghost is a synonym of the label	Ask “Why does this feel the way it does?”
Electric Inflation	Every collision scored Electric	If explainable in 2 seconds, it’s Interesting at best
Vision Without Grounding	Beautiful concept, no Reality Bridge	Never skip Step 5
Homogeneous Fragments	All fragments from one domain	Require at least one external-domain fragment
Forced Collision	Connecting without genuine resonance	Permitted to report “No Electric collisions found”

Table 1: GHOSTY COLLIDER anti-patterns and remedies.

Vision	Market	Competitive	Readiness	External	Overall
Infrastructure Play	Acceleration	Fast Follower	Ready	Open	Go
Experience Layer	Emergence	First Mover	Partial	Opening	Soon
Trust Arbitrage	Pre-emergence	Undefined	Not Ready	Closed	Watch

Table 2: Timing Grid for Case Study B visions.

Dimension	Brainstorming (Control)	GHOSTY COLLIDER (Treatment)
Concept	“Secret crush expressed through dance-pop”	“Concealment as masochistic pleasure — armor that reveals”
Key selection	Convention (“G minor sounds emotional”)	Structural (“F# minor — tonal isolation”)
Modulation purpose	Climactic lift (genre standard)	Narrative surrender
Lyric design unit	Semantic (word meaning)	Phonetic (vowel openness = emotional exposure)
Parameters defined	4	12 (+200%)
Quality score	64/80	75/80 (+17.2%)

Table 3: Music production: brainstorming vs. GHOSTY COLLIDER comparison.

Dimension	Brainstorming (Control)	GHOSTY COLLIDER (Treatment)
Hook strategy	High-energy opening movement	Static gaze (silence as seizure)
Lighting	Multi-source, colorful	Single source, shadow-dominant
Audio	Background music track	No music — ambient sound only
Camera	Dynamic (handheld + crane)	Completely fixed
Structural principle	“More energy = more engagement”	“Stillness as tension = inescapable engagement”

Table 4: Video production: brainstorming vs. GHOSTY COLLIDER comparison.

Dimension	SWOT (Control)	PRECOG PROTOCOL (Treatment)
Temporal structure	None (static snapshot)	4-category action timeline (Now/Soon/Watch/Kill)
Bias detection	None	Systematic (Contrarian View + probabilities)
Timing judgment	Absent	4-axis grid with cross-axis disagreement
Action specificity	“Enter the market” (vague)	Actions with trigger conditions and kill criteria
Contrarian analysis	Not structurally required	3 scenarios quantified (25–60% ranges)

Table 5: Trend forecasting: SWOT vs. PRECOG PROTOCOL comparison.

#	Domain Pairing	Electric	Hit Rate	Visions	Result
1	Urban Agriculture $\times$ Social Design	3	50%	2	Success
2	Elderly Care $\times$ E-sports	3	50%	1	Success
3	Tax Accounting $\times$ Jazz Music	2	33%	1	Success
4	Supply Chain Logistics (homogeneous)	0	0%	0	Failure
5	Dental UX $\times$ Street Photography	2	33%	1	Success
6	Climate Science $\times$ Stand-up Comedy	3	50%	1	Success

Table 6: Batch experiment results (N=8).

Condition	Author Assessment	Blind Evaluation
GHOSTY COLLIDER	75/80	74/80
Brainstorming	64/80	49/80
Gap	11 points	25 points

Table 7: Blind evaluation results for Case Study D.

Metric	Control (Standard)	Treatment (Protocols)
Output correctness	Accurate	Accurate
Structural novelty	Converges to conventions	Qualitatively distinct directions
Parameter specificity	4–6 parameters	12+ parameters (+200%)
Temporal reasoning	Static	Dynamic (timing + triggers)
Bias detection	Not built in	Structural (Contrarian View)
Process reproducibility	Participant-dependent	Protocol-documented

Table 8: Summary of controlled comparisons.

Computational Creativity

Computational creativity studies how machines can generate, evaluate, and collaborate on novel, valuable ideas and artifacts.

See all 114 papers

May 30, 2026

IDEAFix: Evaluation Framework for Creative Defixation Prompting in LLMs

Large language models (LLMs) are increasingly used for tasks involving creative problem solving and idea generation. However, there is a lack of consensus concerning their creative capabilities: some studies report superior performances compared to humans, while others highlight structural limitations such as fixation and the homogenization of outputs. Existing evaluation approaches either rely on narrow, decontextualized tasks that do not capture goal-oriented generation or on broader settings that confound multiple aspects of the creative process, making it difficult to isolate the effects of task formulation, prompting, and evaluation design. Significantly, the role of structured prompting strategies in shaping idea generation remains underexplored. Therefore, we introduce IDEAFix, an evaluation framework for analyzing divergent thinking in open-ended idea generation tasks. We prompt models to generate multiple original solutions to controlled variations of short design scenarios, task attributes, and defixation prompting strategies. This design enables systematic analysis of how structured guidance influences LLMs' idea generation. Our results show that both task formulation and attribute selection significantly affect models' performance, and that simple prompting strategies can boost the originality of solutions. However, we also observe persistent output homogenization across models, confirming inherent limits in their ability to generate diverse solutions. Overall, IDEAFix provides a controlled, extensible framework for studying the mechanisms underlying LLMs' creativity.

F. Carichon, S. Sharma, M. Girard +2

McGill University, Montreal, Canada · Mila, Montreal, Canada · Concordia University, Montreal, Canada

Creativity Assessment LLM Evaluation Computational Creativity

Ask PDF

Sep 3, 2026

Decoupled Analysis-Judging: An Automated Creativity Evaluator Using LLMs in Complex Multi-step Creativity Tasks

Automated evaluation of creativity tasks remains challenging for LLM-as-a-Judge, as LLM is susceptible to biases such as verbosity bias and leniency bias. Such limitations are particularly evident in Contextually-Grounded and Procedurally-Structured Tasks (CGPST), a complex multi-step creativity task where inter-step dependencies, highly subjectivity, and wide scoring ranges lead to more unstable and biased judgments. Existing approaches either rely on task-specific training or directly apply LLM-as-a-Judge, both of which struggle to ensure reliable evaluation under such complexity. To bridge these gaps, we propose CreaEval, an automated creativity evaluator for CGPST that decouples typical LLM-as-a-Judge into analysis and judging. Correspondingly, CreaEval involves two critical phases: Memory-augmented Analysis, a SoT-LLM converts multi-step responses into structured evaluation evidence, incorporating cross-step memory; and Evidence-based Judging, a Judge-LLM uses the extracted evidence for judging without accessing raw responses. Comprehensive experiments show that CreaEval achieves an average performance improvement of 22.74% over the second-best baselines across CGPST and two classic simple creativity tasks, demonstrating its generalizability. The code is available at https://github.com/Jaong/CreaEval.

Xiangyu Wang, Jin Wu, Xiaoyu Li +2

Department of Educational Psychology, East China Normal University · 2Shanghai Institute of Artificial Intelligence for Education, East China Normal University · School of Computer Science and Technology, East China Normal University +2

Creativity Assessment LLM-as-a-Judge Computational Creativity Automated Evaluation

Ask PDF

May 7, 2026

Ex Ante Evaluation of AI-Induced Idea Diversity Collapse

Creative AI systems are typically evaluated at the level of individual utility, yet creative outputs are consumed in populations: an idea loses value when many others produce similar ones. This creates an evaluation blind spot, as AI can improve individual outputs while increasing population-level crowding. We introduce a human-relative framework for benchmarking AI-induced human diversity collapse without requiring human-AI interaction data, providing an ex ante protocol to estimate crowding risk from model-only generations and matched unaided human baselines. By modeling ideas as congestible resources, we show that source-level crowding is identifiable from within-distribution comparisons, yielding an excess-crowding coefficient

Δ

and a human-relative diversity ratio

ρ

. We show that

ρ\ge1

is the no-excess-crowding parity condition and connect

Δ

to an adoption game with exposure-dependent redundancy costs. Across short stories, marketing slogans, and alternative-uses tasks, three frontier LLMs fall below parity across crowding kernels. Estimates stabilize with feasible model-only sample sizes. Importantly, generation-protocol variants show that crowding can be reduced through targeted design, making diversity collapse an actionable, development-time evaluation target for population-aware creative AI.

Nafis Saami Azad, Raiyan Abdul Baten

Bellini College of Artificial Intelligence, Cybersecurity, and Computing University of South Florida Tampa, FL–33620, USA

LLM Evaluation Computational Creativity Generative AI Evaluation

Ask PDF

Abstract

Computational Creativity

Explore similar work

IDEAFix: Evaluation Framework for Creative Defixation Prompting in LLMs

Decoupled Analysis-Judging: An Automated Creativity Evaluator Using LLMs in Complex Multi-step Creativity Tasks

Ex Ante Evaluation of AI-Induced Idea Diversity Collapse