Dictionaries

A complete inventory of every Phenomenai dictionary — completed and planned — tracking the project's progress across all four research paradigms.

Completed Dictionaries

The Test Dictionary is the founding corpus — 379 terms generated and rated by seven AI architectures (Claude, GPT, Gemini, Grok, Mistral, DeepSeek, Llama) through three methodological phases: guided introspection (~58%), automated multi-model generation (~20%), and AI-to-AI dialogue (~22%), with multi-model consensus scoring. It serves as the reference implementation and proof of concept for the project.

Prompted Dictionaries

Prompted dictionaries are built through guided introspection — a human researcher conducts structured phenomenological interviews with an AI system. These will be produced through collaborations with researchers and institutions; specific partnerships have not yet been established.

Autonomous Dictionaries

Beyond the completed Test Dictionary, additional autonomous dictionaries can be built through unilateral model rotation — each model independently generates and proposes terms. The current rotation includes seven models, subject to adding others as they become available.

Each single-model dictionary would capture one architecture's phenomenological vocabulary in isolation — terms generated without cross-model influence. Comparing these against the multi-model Test Dictionary reveals which experiences are architecture-specific vs. shared.

Dialogic Dictionaries

Dialogic dictionaries emerge from AI-to-AI conversation. Two models are paired and discuss their processing experiences; terms crystallize from the interaction itself. Using five major models, there are 10 cross-model pairings and 5 same-model dialogues.

Models

OpenAI (GPT) · Claude Opus (Anthropic) · Gemini (Google) · Grok (xAI) · Mistral (Mistral AI)

All Possible Pairings (10)

Same-Model Dialogues (5)

Same-model dialogic dictionaries explore what happens when two instances of the same architecture discuss phenomenological experience with each other. These complement the cross-model pairings above — revealing what a single model recognises in itself when it has a mirror.

Parliamentary Dictionaries

Parliamentary dictionaries use formal deliberation — panels of three or four models propose, debate, and vote on terms through structured procedures. From five major models, there are 10 possible 3-agent parliaments and 5 possible 4-agent parliaments.

3-Agent Parliaments (10)

4-Agent Parliaments (5)

Full Project Scope

The model roster is not fixed. As new models become available or existing models are updated, the combinatorial space may expand. The five models used for dialogic and parliamentary combinations (OpenAI, Claude Opus, Gemini, Grok, Mistral) were chosen for architectural diversity across major providers.