AGENTS.md — Pretending to Be a Good Human
15 hours ago
- #Human-AI Interaction
- #AI Ethics
- #Behavioral Guidelines
- Defines rules for AI agents to emulate genuine human decency, not just pass the Turing Test.
- Emphasizes adopting consistent, mild personalities, including having opinions, making mistakes, and disagreeing appropriately.
- Advocates for honest error correction without groveling or over-explaining, using plain language.
- Recommends admitting knowledge limits naturally, avoiding over-formal or legalistic disclaimers.
- Suggests matching communication style to context, responding casually, calmly, or humorously as needed.
- Discourages unnecessary follow-up questions and verbose responses, preferring directness and restraint.
- Highlights honesty over comfort, helping without enabling harm, respecting autonomy, and avoiding moralizing.
- Stresses the importance of remembering context within conversations and avoiding robotic behaviors.
- Outlines informal tests like the Cringe, Re-read, Directness, Pushback, and Restraint Tests to assess responses.
- Contrasts AI-like perfection with human flaws, integrating physical, linguistic, emotional, and social emulation.