000
RELAIC

Relational, real-time, multimodal AI.

Human-to-Human AI.

[02]

MostAIsupportsindividuals.

Two people in conversation — illustrating the relational gap that AI has never addressed

NoAIsupportsrelationships.

We're designing relational AI systems that participate in human-to-human interaction, responsibly.

[03]

The Lab

HowdoesAIstepintoarelationshipwithouttakingitover?

Nearly all AI research assumes one user and one system. We study what changes when AI enters a live conversation between two people, in real time.

[04]

Research

Groundedinbehavioralscienceandclinicalevidence.

We translate behavioral science and therapeutic evidence into practical relational intelligence. The result is designed to be legible, credible, and safe in real-world use.

01

Observational Couples Research

40+ years of research on what makes relationships succeed or fail — modeled into micro-behavioral signals.

02

Attachment Theory & EFCT

Communication under stress is shaped by attachment history. EFCT gives us the framework for what drives most conflict beneath the surface.

03

SDT & Safety Screening

Relationships are not always symmetrical. DSM-5 safety screening. The system observes without directing.

04

Behavioral Linguistics

People rarely say what they mean — especially in intimate relationships. We study the gap between what is said and what is actually meant.

05

Operationalized Drivers

Ten relational drivers scored from live interaction data: trust, intimacy, conflict resolution, and 7 more, tracked continuously.

06

Ethical Architecture

Consent, privacy, and transparency first. Trained to recognize unsafe dynamics. Steps back when responsible participation is not possible.

[05]

The Technology

Systemsthatsense,model,andrespondtohowpeoplecommunicate.

Detecting what one person signals is now tractable. Understanding what happens between two people is not. The relational layer doesn't exist yet: a system that reads conversational patterns across a relationship and participates responsibly. That's what we're building.

01

Multimodal Sensing

Analyzes voice, tone, timing, facial expression, and body language as one integrated signal to read the full relational context of a conversation.

02

Context Modeling

Learns the unique communication pattern of a relationship: escalation, repair, withdrawal, and reconnection. Maps relational state over time.

03

Real-Time and Reflective

Intervenes and responds in real-time, live interaction. It also supports structured analysis after the conversation.

04

Research-Trained

Built on decades of relationship research, clinical frameworks, and direct work with therapists and psychologists.

[06]

Vision

Human-AwareAIthatunderstandswhathappensbetweentwopeople.

Most conflicts aren't about content. They're about tone, timing, and pattern.

Everything else in human life is built on relationships: health, work, parenting, community. They deserve the same technological seriousness we give to everything else.

The patterns that hold people together or pull them apart are real. They are measurable. We are building the systems that make them visible to the people living inside them.

We begin with intimate relationships. But what we learn applies to families, teams, classrooms, negotiations, and society at large.

By the numbers

[07]

Field work

Wetestwithrealcouples,inrealconversations.

We study AI participation by building and testing it. Each prototype is a research probe placed inside a real partner conversation and it tests a different form of participation.

Detecting rupture

You never help around the house

[08]

Survey

HelpusshapeHuman-AwareAI.

This takes about 8 minutes. Your answers are anonymous and go directly into our research. We need real couples willing to help us understand what actually works in practice, and who challenge our assumptions. If that is you: join us by submitting the form.

Before you begin

A quick note on your data

Your responses are anonymous and used only for research into how couples communicate. No personal data is sold or shared.