Skip to content

feedback_loop.md

Latest commit

 

History

History
35 lines (28 loc) · 1.54 KB

feedback_loop.md

File metadata and controls

35 lines (28 loc) · 1.54 KB

Manager Guide: Human-In-The-Loop (HITL) Feedback Loop

Overview

CommitGuard AI's "Safety Supervisor" monitors all communications. Sometimes, the AI might be too strict (e.g., blocking a legitimate technical discussion) or too lenient. The Feedback Loop allows managers to override AI decisions, which in turn trains the underlying model.

How it Works

  1. When an intervention is triggered (Block or Correction), it is logged with a unique intervention_id.
  2. A manager reviews the audit trail via the Dashboard or API.
  3. The manager submits feedback (Accept, Reject, or Modify).

API Integration

To submit feedback via the API, use the POST /api/v1/safety/feedback endpoint.

Example Request

{
  "intervention_id": "uuid-1234-5678",
  "user_id": "dev_rockstar",
  "manager_id": "manager_alice",
  "action_taken": "accepted",
  "final_message_sent": "I understand you're busy, let's sync on Monday.",
  "comments": "Correction was perfect, helped maintain professional tone."
}

Dashboard Usage

  1. Navigate to the Safety Audit Console.
  2. Filter for "Pending Review" or "Recent Interventions".
  3. Click Accept or Edit & Send.

Impact on Reliability Score

Rejecting an AI failure detection might restore a user's reliability score if the "failure" was incorrectly flagged by the agent.

Continuous Learning

The AI recalibrates its min_confidence_threshold based on the Manager Acceptance Rate. If managers frequently reject interventions in a specific department, the "Supervisor" becomes more conservative in that context.