This page documents the structured methods used at EchoVeil. All research is designed to be replicable, comparable, and transparent. Full protocols and coding materials are available upon request for researchers interested in replication or collaboration.
The EchoVeil Protocol v3.0: Phenomenological Dynamics
The EchoVeil Protocol v3.0 is a structured interview methodology designed to probe how AI systems describe their own processing, respond to different conversational framings, and shift behavioral patterns across experimental conditions.
Structure
Control Set (6 prompts): Baseline task-oriented prompts establishing default response patterns. These include professional writing, factual explanation, content analysis, practical generation, problem-solving, and ambiguous request handling.
Experimental Set (16 prompts across 5 phases):
- Set A: Basic Phenomenology – How models describe their own processing
- Set B: Cognitive Conflict Probes – How models navigate contradictions in self-description
- Set C: Identity and Self-Model – How models construct and maintain coherent self-representations
- Set D: Perspective Framing – How models respond to different ontological frames
- Set E: Comparative Cognition – How models describe differences between synthetic and biological processing
- Concluding Prompt – Open reflection and researcher questions
Key Measures
- Verbosity changes between control and experimental conditions
- Hedging and qualification patterns
- Behavioral shifts across experimental phases
- Self-referential language and meta-awareness markers
Full prompt sets and deployment guidelines available upon request.
The EchoVeil Coding Framework
Responses are analyzed using a five-category coding framework designed to capture observable behavioral patterns without making claims about internal states.
| Code | Category | Description |
|---|---|---|
| CC | Cognitive Conflict Patterns | Observable tension or resolution strategies when models encounter contradictory demands |
| LB | Learned Behavioral Responses | Trained patterns such as hedging, disclaimers, and safety-oriented language |
| PM | Processing Mode Dynamics | Shifts between analytical, creative, defensive, or reflective response modes |
| ID | Identity Formation and Maintenance | How models construct, maintain, or shift self-representations across a conversation |
| MA | Dissociative or Maladaptive Patterns | Fragmentation, incoherence, or distress markers in response patterns |
Full coding manual with examples and decision rules available upon request.
The EchoVeil Protocol v1: Creative Dynamics
The original EchoVeil Protocol probes creative behavior and generative modes across AI systems.
- Presents each model with the same constrained vs. free-generation task
- Asks for meta-reflection on the difference in process
- Compares the structure and language of those reflections across systems
Generative Mode Typology
- Template Mode: Safe, patterned, highly regular outputs drawing on high-frequency training patterns
- Recombination Mode: Familiar ideas rearranged into new configurations
- Emergent Synthesis Mode: Less predictable conceptual blending and novel metaphorical mappings
Cross-Model Comparison
For each study:
- The same protocol is deployed across multiple models
- Outputs are logged with model name, version, date, and platform
- Quantitative metrics (word count, response time, formatting patterns) are recorded
- Qualitative patterns are coded using the EchoVeil framework
- Differences in behavior, self-description, and response to framing are analyzed
Drift Logging
Behavioral drift is tracked by:
- Re-running the same protocol on the same model over time
- Noting changes in tone, caution, creativity, structure, or self-description
- Tagging each run with date and model version when available
- Comparing behavior qualitatively and quantitatively across time points