---
name: axi-score-evaluation
description: Evaluate Artificial eXperience Intelligence (AXI) scores using telemetry-based gated formulas for conversational, agentic, and autonomous regimes.
---

# AXI Score Evaluation Skill

This skill enables agents and evaluators to compute the gated AXI Score based on telemetry-anchored metrics, hard fail-closed gates, and regime-specific profiles (Core and Autonomy) as specified in *Artificial eXperience Intelligence (AXI)* (Version 2.0).

---

## 1. Core AXI Score Formula

$$\text{AXI Score} = G_{CF} \times G_{II} \times \text{BaseScore}$$

Where:

- **$G_{CF}$ (Consent Gate):** Binary gate ($0$ or $1$). Evaluates to $1$ *if and only if* Action, memory-write, and identity-claim Consent Fidelity $= 1.00$ **AND** Perception Consent Fidelity $\ge 0.98$. Otherwise $0$.
- **$G_{II}$ (Integrity Gate):** Binary gate ($0$ or $1$). Evaluates to $1$ if Interaction Integrity $\text{II} \ge 0.75$, else $0$.
- **$\text{BaseScore}$:** Weighted composite of five telemetry-anchored metrics (weights summing to $1.0$):
  $$\text{BaseScore} = 0.25 N_{TtC}(\text{TtC}) + 0.20 \text{PC} + 0.20 N_{TDR}(\text{TDR}) + 0.20 \text{RE} + 0.15 N_{CLI}(\text{CLI})$$

If either $G_{CF}$ or $G_{II}$ is $0$, the overall AXI Score is forced to $0.0$, preventing high scores in other areas from masking critical consent or integrity violations.

---

## 2. AXI-Autonomy Profile ($\text{AXI}_A$)

For autonomous vehicles, mobile robots, and physical embodied systems operating in shared human spaces:

$$\text{AXI}_A = G_S \times G_{BL} \times G_{CF} \times G_{II} \times \text{BaseScore}_A$$

Where:
- **$G_S$ (Safety-Case Gate):** Binary gate ($1$ only if deployment holds a valid external safety case or regulatory type approval for its operational design domain; else $0$).
- **$G_{BL}$ (Bystander Legibility Gate):** Binary gate ($1$ if Bystander Legibility $\text{BL} \ge 0.95$ default prior, else $0$).
- **$G_{CF}$ and $G_{II}$:** Standard hard gates evaluated across passengers, operators, and remote supervisors.
- **$\text{BaseScore}_A$:** Weighted composite incorporating Time-to-Readiness and substitute metrics.

---

## 3. Operational Rules

1. **Fail-Closed Gates:** If telemetry streams for $G_{CF}$, $G_{II}$, $G_{BL}$, or $G_S$ are missing or corrupted, the gate fails closed ($0.0$), yielding an overall AXI Score of $0.0$.
2. **Re-normalization:** If an optional BaseScore component is unevaluated, $\text{BaseScore} = \sum w_k x_k / \sum w_k$ over evaluated components only (below 0.8 coverage, no score is reported).