Mimetic Models: Ethical Implications of AI that Acts Like You

Mimetic models create new ethical risks beyond traditional AI systems because they're specifically designed to replicate individual people's decision-making and behavior patterns, rather than making generic predictions. This personalization raises distinct concerns about consent, identity, and misuse that earlier AI approaches didn't present.

Multiple stakeholders are affected differently by mimetic models, including the person being modeled, the operator deploying the system, and those interacting with it. The framework highlights that ethical analysis must consider impacts across all these groups rather than focusing on just one perspective.

Mimetic models are designed for ongoing interaction, not just one-time outputs, which means they can accumulate influence and create ongoing relationships with users. This interactive nature amplifies potential harms compared to static, pre-computed AI systems.

Current applications span high-stakes domains including game-playing, text generation, and artistic expression, suggesting mimetic models are already being deployed across diverse contexts where they could significantly impact people's decisions and experiences.

A comprehensive ethical framework is needed urgently because the paper identifies this as an "emerging theme" in AI research, indicating that governance and ethical guidelines haven't yet caught up with the technology's development and deployment.