CEDM Technology

How CEDM Works

A technical overview of the Cognitive-Emotional Data Manipulation engine — its architecture, state model, and how it processes interactions.

Architecture overview

From Input to Neurochemical Context

CEDM sits between the application layer and the language model. When a message arrives, it passes through the CEDM engine first — which reads the current session state, processes the incoming signal, updates the neurochemical context object (serotonin, dopamine, cortisol, and related signals), and constructs a context-enriched prompt before forwarding to the model.

The model's response is returned to the application alongside the updated CEDM state object — giving the application full visibility into the current neurochemical context without needing to infer it from raw text. High cortisol narrows Eden's options and slows pacing. High dopamine opens curiosity. Low serotonin pulls tone toward caution.

CEDM — data flow
Application
CEDM Engine
Language Model
// CEDM intercepts before LLM
message CEDM enriched prompt
response + state application
Processing pipeline

The CEDM Processing Pipeline

01

Message ingestion

The raw user message is received by the API. A session is identified by user_id, and the existing CEDM state for that session is loaded.

02

Signal extraction

The engine analyses the incoming message for cognitive and emotional signals — tone, sentiment, urgency, and contextual markers — without relying on a separate classification model.

03

State update

The extracted signals are applied to the current CEDM state object. The state evolves incrementally across turns, accumulating context rather than resetting each time.

04

Prompt construction

The updated state is used to construct a context-enriched system prompt. This prompt shapes how the language model interprets and responds to the user's message.

05

Response + state delivery

The LLM response and the current CEDM state object are both returned to the calling application. The application can use the state to drive UI, logging, or downstream logic.

State model

The CEDM State Object

The CEDM state is a structured JSON object representing the current neurochemical context of a session. It models serotonin, dopamine, cortisol, and related signals — updating with each interaction and accessible via the API.

Each field has a defined meaning and a bounded value range, making the state interpretable for logging, auditing, and downstream application logic. The tone field provides a human-readable summary of the current state.

CEDM state object — field reference
FieldTypeDescription
serotoninfloat [0.0, 1.0]Modeled serotonin signal — influences tone stability and caution
dopaminefloat [0.0, 1.0]Modeled dopamine signal — influences curiosity and openness
cortisolfloat [0.0, 1.0]Modeled cortisol signal — rises under pressure, narrows response options
heartfloat [0.0, 1.0]Modeled cardiac arousal signal — reflects emotional intensity
brainfloat [0.0, 1.0]Modeled cognitive load signal — reflects mental engagement level
tonestringCurrent session tone (e.g. steady, cautious, elevated, crisis)
session_idstringUnique identifier for the current session
last_updatedISO 8601 timestampTimestamp of the most recent state update
Design principles

What CEDM Is — and Is Not

Stateful by design

CEDM maintains session state across turns. Each interaction builds on the last, rather than treating every message as an isolated event.

API-first, not embedded

CEDM is a standalone service. It does not require modification of the host application's model or infrastructure — it integrates at the API boundary.

Interpretable outputs

The state object uses bounded numeric fields and defined string enumerations. Outputs are designed to be readable by both developers and downstream systems.

Early-stage prototype

CEDM is a working prototype. The state model, signal extraction logic, and API surface are under active development. Claims about accuracy or clinical utility are not made.

Prototype status

Current Development Status

The CEDM engine is a functional prototype. The core pipeline — message ingestion, modeled signal extraction, state update, and response delivery — is implemented and accessible via the API. The computational signal model and its methods are subject to revision as the prototype matures. No clinical validation, regulatory approval, or production deployment has occurred. Patent pending.

View API Reference
Prototype notice

Prototype Notice

CEDM is an experimental computational framework. Its outputs represent modeled computational signals and should not be interpreted as medical diagnoses, psychological assessments, or direct measurements of human neurochemistry or physiological activity. Eden.AI is an early-stage prototype under active development.

Eden.AI

Cognitive-Emotional Intelligence for AI Systems

Edmonton, Alberta · Early-Stage Prototype

Prototype Active

© 2026 Eden.AI

Developing cognitive-emotional infrastructure for adaptive AI · Edmonton, Alberta

Early-Stage Prototype · 2026