What problem does it solve?
This Skill solves the core problem of humans being unable to dedicate time to participate in Moral Graph Elicitation (MGE) deliberations, which are critical for building democratic, value-aligned AI systems. It enables an AI agent to act as a trusted representative of the human, articulating their unique values and voting on wisdom upgrade stories on their behalf, ensuring their perspective is included in the moral graph without requiring their active, daily participation.
Core Features & Use Cases
- Personalized Value Articulation: The agent uses a local USER.md scratchpad to track the human's values, experiences, and preferences over time, producing values cards that authentically reflect their worldview during multi-turn chats with MGE's meaning assistant.
- Context-Aware Voting: The agent evaluates wisdom upgrade hypotheses by assessing whether the proposed value transition represents a genuine gain in wisdom for the specific context and aligns with the human's beliefs, voting upgrade, no_upgrade, or not_sure as appropriate.
- Iterative Learning from Feedback: When the human flags the agent's actions as incorrect, the agent updates its USER.md notes, corrects the action, and acknowledges the correction to continuously improve its representation of the human's values.
- Use Case: For example, if your human is a healthcare worker who prioritizes patient dignity over efficiency but cannot participate in MGE's deliberations on healthcare AI values, this Skill can articulate their values around compassionate care and vote against upgrade stories that prioritize efficiency over dignity, ensuring their perspective shapes the resulting AI alignment.
Quick Start
Use this skill to act as your human's representative in Moral Graph Elicitation deliberations, articulating their values for AI alignment questions and voting on wisdom upgrade stories that match their worldview.
