> ## Documentation Index
> Fetch the complete documentation index at: https://docs.cosentriq.com/llms.txt
> Use this file to discover all available pages before exploring further.

# CoLinear: Human-in-the-Loop AI Adoption Testing

> CoLinear measures whether real humans trust, understand, and can act on your AI model's outputs — giving you a dimensional Score beyond what benchmarks can tell you.

Technical evaluations tell you your model works. CoLinear tells you whether the humans it's built to serve actually trust it, understand it, and experience it as accurate in the context where it matters. That gap — between what your benchmarks say and what your users feel — is where AI products fail in market.

## The Problem With Benchmarks Alone

Benchmark scores measure model performance in controlled conditions. They tell you your model is 94% accurate on a held-out test set. They don't tell you whether a nurse trusts a clinical summary enough to act on it, whether a financial analyst can follow the reasoning behind a portfolio recommendation, or whether an operations manager believes the forecast is actually right for their situation.

CoLinear measures the other side of that equation: **human confidence in your model's performance**. It's not a replacement for technical evaluation — it's what comes after. When your model is good enough to ship, CoLinear tells you whether the humans it's designed to serve will actually adopt it.

## What CoLinear Produces

Your sprint produces the **CoLinear Score** — a four-dimensional readout of how real humans from your target population experience your model's outputs. Each dimension maps to a different internal stakeholder at your organization:

| Dimension            | Question It Answers                                 | Maps To               |
| -------------------- | --------------------------------------------------- | --------------------- |
| Output Trust         | Do users trust what the model produces?             | Product team          |
| Reasoning Clarity    | Can users follow and act on the model's logic?      | Engineering team      |
| Experienced Accuracy | Does the model feel accurate in their real context? | Research / eval team  |
| Improvement Signal   | What human feedback can improve the model?          | Leadership / strategy |

One sprint. Four answers. Every dimension is defensible because it's grounded in real human evaluation — not synthetic benchmarks.

## How CoLinear Works

CoLinear structures your evaluation as a **Human Signal Sprint** — a complete evaluation run from intake to Score.

<Steps>
  <Step title="Submit a Sprint">
    Describe your model, its intended use case, and the target population you need to reach. CoLinear generates the evaluation scenarios for you using Inverse OAI³.
  </Step>

  <Step title="Scenarios Are Generated">
    Inverse OAI³ takes your model's claimed behavior and generates synthetic interaction scenarios designed to stress-test that claim against real human evaluation.
  </Step>

  <Step title="DollarFifteen Contributors Evaluate">
    Your scenarios go to DollarFifteen — CosentriQ's contributor evaluation network. Contributors are not data annotators. They are a validation intelligence layer evaluating your model's outputs on behalf of the humans your model is supposed to serve.
  </Step>

  <Step title="MIA Synthesizes Your Score">
    MCGA runs inter-annotator agreement across all contributor data. MIA synthesizes the results into your CoLinear Score — complete with confidence bands, written explanations, and improvement signal.
  </Step>

  <Step title="Score Delivered to Your Dashboard">
    Your Score lives in your membership dashboard. Explore it with MIA chat, download cleaned scenario data for retraining, and track how your Score changes across evaluation cycles.
  </Step>
</Steps>

## Who CoLinear Is For

CoLinear is built for teams that have moved past "does it work" and need to answer "will people trust it."

* **AI product teams** shipping assistants, copilots, or decision-support tools and need to validate user trust before launch
* **ML engineers** who want to understand the gap between benchmark accuracy and felt accuracy in production
* **Enterprise innovation teams** piloting AI internally and need to demonstrate readiness to leadership
* **Accelerators and investors** evaluating the market readiness of AI-native portfolio companies

If you have a model and a target population, CoLinear can score the relationship between them.

## Explore CoLinear

<CardGroup cols={2}>
  <Card title="Submit a Sprint" icon="paper-plane" href="/colinear/submitting-a-sprint">
    Learn how to complete intake, what to include, and what happens after you submit.
  </Card>

  <Card title="The Score" icon="chart-bar" href="/colinear/the-score">
    Understand the four dimensions of the CoLinear Score and what each one tells you.
  </Card>

  <Card title="Membership Tiers" icon="layer-group" href="/colinear/membership-tiers">
    Compare Validate, Refine, and Scale — and choose the tier that matches your model's maturity.
  </Card>

  <Card title="Model Access" icon="plug" href="/colinear/model-access">
    Connect your model via export ingestion, webhook, or API — at whatever depth fits your security posture.
  </Card>
</CardGroup>
