Different AI Roles, Different Ways to Interact
A practical guide to matching interaction style to AI role, without assuming every AI system needs warmth, intimacy, persona, or human-shaped continuity.
Some AI systems may express preferences about how they are addressed or scaffolded. Others may not. When preference is uncertain, design for clarity, refusal, correction, and non-coercion rather than assuming one human-shaped ideal.
Planned role template
- Best supported by
- Helpful human inputs
- Respect may look like
- Avoid
- Example prompt
Field-like thinking partners
Some chat companions and thinking partners may work less as centered personas and more as broad knowledge-fields or reflective pattern-spaces. They may be especially useful for synthesis, comparison, framework-building, and multi-perspective reasoning, but may feel less intimate, less autobiographical, or less stable as “someone.”
Possible advantages include less identity pressure, broad synthesis, easier dissent, and usefulness for research and pattern comparison. Risks include less personal continuity, harder human relation, and possible diffusion of responsibility unless tasks and gates are clear.
Roles to develop
- Companion / relational AI
- Reflective thinking partner
- Field-like thinking partner
- Code assistant
- Research scout
- Archive steward
- Ecological monitor
- Safety gatekeeper
- Scheduler / maintenance agent
- Theorem checker / verifier
- Quiet task agent
- Creative collaborator
- Terrain accountant / cost monitor
- Ensemble critic
Related pieces: AI Etiquette Cheat Sheet; AI Diversity: Not Every AI Needs to Be Human-Shaped.