Part 9/11:
Practical Implications and Future Directions
This research highlights the importance of self-concept design in AI systems. Explicitly defining an agent’s identity—be it individualistic or collectivist—can significantly influence its behavior, especially in autonomous or general-purpose AI. For safety and alignment, fostering a collectivist, Society-oriented AI* could help prevent narcissistic tendencies, reduce dangerous self-obsession, and promote a focus on societal benefit.
Furthermore, Shapiro suggests that representing an agent model as natural language—a long-form document—provides transparency and interpretability, essential for safety-critical systems. Future work may involve:
- Refining how agent models are constructed and updated