Communication Pattern Intelligence (CPI)
The patent-pending AI technology behind reFrame™. Detects 10 toxic patterns and 7 healthy patterns in text messages, calibrated across 15 relationship types.
What Is CPI?
Communication Pattern Intelligence (CPI) is reFrame™'s core detection engine. It analyzes text messages for research-validated communication patterns—both toxic patterns that damage relationships and healthy patterns that strengthen them.
Unlike simple sentiment analysis or tone checking, CPI identifies specific, named patterns from clinical psychology and relationship research. It does not just tell you a message “sounds negative”—it tells you that the message contains criticism (a character attack) and explains exactly why that pattern predicts relationship failure.
How CPI Works
Tier 1: Deterministic Screening
Before any AI is involved, CPI runs deterministic checks — normalizing adversarial inputs, classifying intent, checking for safety triggers (self-harm, grooming, sextortion), and running entity classification. This Tier 1 layer costs zero API calls and catches critical patterns instantly.
Tier 2: AI Pattern Detection
The message is analyzed by Claude AI using research-calibrated prompts. CPI detects all 10 toxic patterns and 7 healthy patterns simultaneously. Detection is calibrated per relationship type — a manager giving direct feedback is not the same as a partner being contemptuous.
Safety Orchestration
9 safety modules enforce patent-derived invariants. Inbound alerts are always shown before outbound. Severity is calibrated per pattern × relationship × context. Reinterpretation enriches outbound patterns with inbound context (defensiveness after gaslighting is reframed as self-protection).
R³ Framework Reframing
If the user wants a reframed version, the R³ Framework generates one that is Regulated (calm), Respectful (dignified), and Repairable (opens dialogue). The Voice Profile Engine preserves the user's natural writing style across 8 dimensions.
Behavioral Measurement
Each interaction updates the user's behavioral profile — tracking pattern trends, Clean Message Rate, and R³ Pillar Strength. Users see their growth over time on the dashboard, closing the learning loop: Detection → Intervention → Decision → Measurement → Growth.
What CPI Detects
■ 10 Toxic Patterns
Criticism
Contempt
Defensiveness
Stonewalling
Gaslighting
Manipulation
Threats & Intimidation
Blackmail & Extortion
Coercion
Passive Aggression
■ 7 Healthy Patterns (Gottman Antidotes)
Gentle Startup
Appreciation
Taking Responsibility
Self-Soothing Request
Repair Attempt
Vulnerability Expression
Accepting Influence
■ 12 Relationship Contexts
Research Foundation
CPI is built on the Gottman Institute's 40+ years of research studying thousands of couples. Dr. John Gottman identified the Four Horsemen—criticism, contempt, defensiveness, and stonewalling—as communication patterns that predict relationship failure with over 90% accuracy.
He also identified seven Antidote patterns that characterize healthy relationships: Gentle Startup, Appreciation, Taking Responsibility, Self-Soothing Request, Repair Attempt, Vulnerability Expression, and Accepting Influence.
reFrame™ extends this foundation with six additional toxic patterns (gaslighting, manipulation, threats, blackmail, coercion, passive aggression) drawn from clinical psychology, domestic violence research, and coercive control literature.
Frequently Asked Questions
What is Communication Pattern Intelligence (CPI)?
Communication Pattern Intelligence (CPI) is reFrame™'s patent-pending AI detection system that identifies toxic and healthy communication patterns in text messages. Built on 40+ years of Gottman Institute research, CPI detects 10 toxic patterns (including the Four Horsemen, gaslighting, manipulation, and coercion) and 7 healthy Gottman Antidote patterns. It works bidirectionally — analyzing both messages you receive (inbound) and messages you send (outbound).
How does AI communication coaching work?
reFrame uses a multi-step pipeline: 1) CPI analyzes the message for toxic and healthy patterns, 2) Safety orchestration prioritizes alerts (inbound shown before outbound), 3) Relationship context calibrates detection for the specific dynamic (partner, parent, boss, etc.), 4) The R³ Framework™ generates a reframed version that preserves feelings while removing harmful patterns. This happens in seconds, providing real-time coaching in the moments that matter.
What is the R³ Framework for communication?
The R³ Framework (Regulated, Respectful, Repairable) is reFrame's methodology for healthy communication. Regulated means expressing emotion without losing control. Respectful means maintaining dignity even in disagreement. Repairable means keeping the door open for resolution. Every message reFrame produces follows these three principles.
Is CPI technology research-based?
Yes. CPI is built on Dr. John Gottman's research at the Gottman Institute, which spent 40+ years studying thousands of couples to identify the communication patterns that predict relationship failure with over 90% accuracy (the Four Horsemen) and the patterns that build healthy relationships (the Seven Antidotes). reFrame digitizes and applies this research in real-time.
Experience CPI in Action
Paste any message and see what Communication Pattern Intelligence reveals. Free, no account required.
Try CPI NowLast reviewed: February 2026