Cognitive Scoring Matrix

Every thinking tool and recipe in the Brainstorm Reactor is indexed along a 3-dimensional cognitive taxonomy. Rather than relying on generic keyword search, the platform mathematically quantifies how each method thinks — enabling real-time method recommendation and model-slot matching.

In the left sidebar (Core Console), this framework manifests as an interactive 2D Targeting Matrix.

1. The Three Cognitive Dimensions

Each method in the library is scored on a scale from 0 to 100 across three fundamental axes:

Dimension	Axis Focus	What It Measures	Score 100 Example
Structure	Operational Rigor	Explicit procedural steps, strict XML tags, output schemas, and bounded rules.	TRIZ Contradiction Matrix, Formal FMEA
Divergent	Lateral Exploration	Creative breadth, unexpected analogies, paradigm disruption, and random stimuli.	Brian Eno Oblique Strategies, SCAMPER
Convergent	Analytical Synthesis	Multi-step deductive logic, epistemic validation, contradiction pruning, and trade-off evaluation.	First Principles Audit, Socratic Inquest

2. The 2D Targeting Plane (Core Console)

When navigating the Core Console in the sidebar, you interact with a continuous 2D plane:

Horizontal Axis: Divergence (Exploration & Novelty)
Vertical Axis: Convergence (Analytical Depth & Synthesis)

       Convergence (Logic & Synthesis)
                      ▲
                      │     [Convergent / Analytical]
                      │     e.g., First Principles, SWOT
                      │
   ───────────────────┼───────────────────► Divergence (Creativity)
                      │     [Divergent / Lateral]
                      │     e.g., SCAMPER, Provocation
                      │
                      ▼

Real-Time Fit Calculation

As you move your cursor across the matrix:

The engine computes the normalized Euclidean distance between your cursor position and the 3D coordinate vector of every available method.
Methods with the closest match receive the highest Fit Score (%) and are dynamically sorted to the top of your recommendations list.
This allows you to visually dial in your exact cognitive intent (e.g. "80% Divergent Ideation + 30% Critical Review") without formulating complex search queries.

3. "Optimized for AI" Indicator

Methods with a Structure Score $\geq 95$ carry the dedicated Optimized for AI indicator.

These methods feature mathematically hardened system prompts, explicit error-handling tags, and zero-ambiguity step definitions. When applied to reasoning models (e.g., Claude Opus, DeepSeek R1, GPT-o-series), they yield maximum reliability and near-zero hallucination rates.

4. Model-Aware Slot Routing

The Scoring Matrix also informs how the Brainstorm Reactor assigns backend LLMs in multi-agent recipes:

High Convergent Methods ($\geq 70$): Automatically routed to SLOT_LOGIC (Thinking / Reasoning Models with extended inference budgets).
High Divergent Methods ($\geq 70$): Automatically routed to SLOT_CREATIVE (High-throughput, laterally associative models with elevated temperature profiles).
Structure & Fact Extraction: Routed to SLOT_SEARCH (Fact-grounded research agents with live web-retrieval capabilities).

Core Interface & The Pool — Navigating the sidebar and console
Direct Mode — Applying methods to single prompts
Recipe Builder — Chaining scored methods into multi-agent workflows